Multi-scale in-silico modelling to unveil structural requirements for DNA-PK inhibitors as radiosensitizers and MolSHAP based design of novel ligands

Multi-scale in-silico modelling to unveil structural requirements for DNA-PK inhibitors as radiosensitizers and MolSHAP based design of novel ligands | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multi-scale in-silico modelling to unveil structural requirements for DNA-PK inhibitors as radiosensitizers and MolSHAP based design of novel ligands Soumya Mitra, Rakesh Kumar Dolai, Nilanjan Ghosh, Subhash C. Mandal, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7479073/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 22 Dec, 2025 Read the published version in Journal of Computer-Aided Molecular Design → Version 1 posted 7 You are reading this latest preprint version Abstract Radiosensitizers are agents that make tumour cells more sensitive to radiation therapy. One key mechanism involves inhibition of the DNA-dependent protein kinase (DNA-PK), an enzyme crucial for repairing DNA double-strand breaks in mammalian cells. Suppression of the DNA-PK enzyme compromises the double-strand break repairs to amplify the radiation induced toxicity among the tumour cells. In this study, 73 6‑Anilino Imidazo[4,5‑c]pyridin-2-one derivatives were curated as potent DNA-PK inhibitors and subjected them to 2D -and 3D-Quantitative Structure Activity Relationship (QSAR) analyses to explore their structural requirements. Apart from conventional methodology, we implemented newly developed MolSHAP analyses for R-group analyses. Significant information regarding structural requirements were retrieved from each of these cheminformatic analyses. Additionally, to understand the interaction between the ligands and the DNA-PK receptor, molecular dynamics (MD) simulation analysis of 100ns were carried out for the most and the least potent compounds among the dataset. The findings indicated H-bond and π-π interactions to be the key factors for binding interactions. Furthermore, novel ligands were designed through the MolSHAP tool and were validated through the chemometric model developed in this investigation. The designed compound exhibited favourable predicted activity and replicated key interaction profiles of the co-crystallized bound ligand in MD simulations. The investigation was carried out through open-access tools to safeguard reproducibility and accessibility among researchers. Radiosensitizers DNA-dependent Protein kinase (DNA-PK) QSAR molecular dynamics simulation MolSHAP Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. INTRODUCTION Cancer remains one of the most complicated diseases to appear during the course of history with no definite therapeutic strategies to succeed. With over 1.4 million new cases detected in India alone, therapies related to the management of cancer is highly dependent on radiotherapy. Radiotherapy (RT) has been one of the most commonly used approaches in the management of different types of cancers with around 52.6% of the total population of cancer patients undergoing RT at different stages of the disease [ 1 , 2 ]. RT, in combination with chemotherapy, along with radiosensitizers has become the typical approach for treating head and neck small cell carcinoma [ 3 ]. While the strategy is beneficial in cases with detection of cancer, this therapeutic approach also presents with high occurrences of toxicity towards the unaffected cells and tissue [ 4 ]. To combat the adverse effects associated with radiotherapy, several investigations have been dedicated to enhance the efficacy of the RT with minimal exposure of the patients to the radiation. One such strategy involves the introduction of radiosensitizers which are chemical compounds capable of increasing the efficiency of the RT. Radiosensitizers are effectively compounds which make the tumor cells more susceptible towards RT. Radiotherapy deals with ionizing radiation which can damage the DNA in the cancerous cells in irreversible manner. RT produces free radical, especially hydroxyl radicals, which promote breaking of the DNA double strand. The oxygen then reacts with the hydroxyl free radical to give rise to peroxide radicals which causes a more permanent damage towards the DNA strands [ 5 ]. In case of tumors, they are often seen in regions with low oxygen content leading to hypoxic tumours. Having scarcity of oxygen often leads to resistance towards RT which is caused by the inability to form peroxide free radicals. To overcome the RT associated resistance, investigators have resorted to developing compounds like nitroimidazole, which can act as substitute for oxygen in those low oxygen environments and cause irreversible DNA damage[ 6 – 9 ]. Another strategy to counter the RT resistance is by developing protein kinase inhibitors that primarily stops the hypoxic cells from recovering from the DNA damage. In recent years, more intensive investigations have been carried out which focussed on the pathways that lead to DNA damage repair through development of modern radiosensitizers [ 10 , 11 ]. RT has been responsible for creating double strand breaks (DSB) while inducing cytotoxic effects on the cells. Resistance towards RT can be indicated by the ability of the DNA to repair itself; primarily by two pathways; which include non-homologous end joining (NHEJ) and homologous recombination (HR) [ 12 ]. However, the NHEJ plays the predominant role as it increases radiosensitivity of cells by mutating the NHEJ genes [ 11 , 13 ]. Within the NHEJ pathway, the Ku heterodimers rapidly locate the DSB and attach itself to the DNA terminal to prevent any further nucleolytic damage [ 14 ]. Irrespective of the structural configuration of the DNA, the Ku dimers show high binding affinity to the damaged ends. Due to its ring-like shape (toroidal) it can easily attach itself to the DNA terminal and start recruiting DNA-PK catalytic subunits (DNA-PKcs) along with the nuclease Artemis. Artemis is responsible for trimming the damaged ends of the DNA and together with the DNA-PKcs, they form a DNA-PK holoenzyme. The association of the Ku with DNA-PKcs increases the stability of the Ku-DNA complex and promotes a conformational shift in DNA-PKcs which activates the kinase function through autophosphorylation at multiple regulatory sites. Once activated, DNA-PKcs induces phosphorylation at the C-terminal of the autoinhibitory segment of Artemis, enabling to perform the nuclease activity Following end processing, the DNA-PKcs recruit gap-filling polymerases from the Pol X family, specifically DNA polymerase λ and DNA polymerase µ. These enzymes restore the missing nucleotides, following which the DNA ligase IV-XRCC4-XLF complex completes the ligation to facilitate DNA repair [ 15 – 22 ]. This loss of DNA-PK functions severely sensitises the DSBs [ 23 , 24 ], that makes DNA-PK as major drug target [ 11 , 25 , 26 ]. Recently it was confirmed that the DNA-PK offers the higher radio sensitization than inhibitors of PARP-1 (poly (ADP-ribose) polymerase-1) in HNSCC (Head and neck squamous cell carcinoma) cells [ 27 ]. The limited resolution of DNA-PK structural data first made it difficult to find strong and specific DNA-PK inhibitors [ 28 – 31 ], and dependent on the homology model [ 32 , 33 ] based on the phosphoinositide 3-kinase (PI3K) enzymes. Because of this, initial inhibitors like dactosilib, NU7441, KU0060648, and IC87361 show lack of precise selectivity to DNA-PK, they affected not only DNA-PK but also PI3K and other related enzymes in the PIKK (PI3K-related kinase) family [ 34 – 39 ]. This lack of selectivity led to unwanted adverse effect or toxicities in preclinical studies, limiting their therapeutic potency [ 40 ]. In recent time, more selective inhibitors to DNA-PK such as AZD7684, NU5455, Pepostertib (M3814), and 31t are identified and most of these are characterized at higher resolution in structural studies [ 41 – 47 ]. Two of them, Peposertib with RT as well as AZD7648 in combination with pegylated doxorubicin or Olaparib, have progressed to clinical trials. In the current work, we collected a data set containing 73 6‑Anilino Imidazo [4,5‑c] pyridin-2-one’s derivatives that were recently reported as novel DNA-PK inhibitors and performed a range of in silico modelling analyses to understand the structural requirements in order to ensure that more robust design strategy may be adopted with such information for obtaining DNA-PK inhibitors with greater potency [ 48 ]. 2. MATERIALS AND METHODS 2.1 Dataset curation and preparation In search of potent DNA-PK inhibitors, Hong et al. designed a series of novel 6Anilino Imidazo[4,5c] pyridin-2-one derivatives by deploying scaffold hopping techniques on an established PI3K and PIKK inhibitor Dactolisib [48]. The group developed new inhibitors and evaluated their inhibitory potential against the DNA-PK enzyme using the HotSpot assay platform. The nanomolar IC 50 values reported for the 73 compounds were subsequently transformed into pIC 50 (= ‒log 10 (IC 50 /10 9 ) values. The SMILES notations of the dataset compounds are collected from the article of Hong et al [ 48 ]. The dataset was prepared by converting the curated SMILES into their canonical forms via RDKit (available at https://www.rdkit.org/ ) and were converted into .sdf files through Discovery Studio Visualizer tool (available at https://discover.3ds.com/discovery-studio-visualizer-download ). This was followed by incorporation of the biological activity in terms of pIC 50 as the response variables required to develop QSAR models. Once the dataset was ready (Table S1 of the Supplementary information), the ligands of the dataset were subjected to structural standardization by the CDK standardizer. The standardization process entails adding explicit hydrogen atoms, performing aromatization, cleaning the structure in both 2D and 3D, neutralizing, and removing any salts. 2.2 2D-QSAR modelling 2.2.1 Descriptor calculation For descriptor calculation, the open-access webserver OCHEM ( https://ochem.eu/login/show.do ) was utilized where 73 dataset ligands were fed to compute alvaDesc descriptors ( https://www.alvascience.com/alvadesc-descriptors/ ) in . sdf format [ 49 , 50 ]. The molecular structures of the 6anilino imidazo[4,5c] pyridin-2-one derivatives were carried out by the RDKit module of the webserver. The descriptor calculation took place in two steps, wherein the first step included computation of all the alvaDesc descriptors and the next step included computing only those alvaDesc descriptors which have better interpretability. The details of the interpretable descriptors have been provided in one of our previous works [51].The descriptor files (. csv ) were prepared by adding the identifier of dataset ligands and their respective pIC 50 values as response variables. 2.2.2 Dataset division and model set-up As both the datasets with all AlvaDesc descriptors and interpretable descriptors were developed, they were segregated into training and test (or prediction) sets by using SFS-QSAR-tool_v2 (available at https://github.com/ncordeirfcup/SFS-QSAR-tool_v2 ) [ 51 ]. The datasets were divided while deploying activity sorting 4, and the ratio between training and prediction set were kept at 4:1. In regression analysis, feature selection helps in identifying the relevant features during model development [ 52 , 53 ]. Two distinct feature selection modules were utilized for setting up the 2D-QSAR models; (i) sequential forward selection and (ii) genetic algorithm. The first method SFS works by iteratively adding the most significant feature that improves model performance until a stopping criterion is met. On the other hand, GA is a heuristic search inspired by natural selection [ 51 ]. Sequential forward selection (SFS) is a non-stochastic greedy algorithm which acts by selecting the optimum features in a dataset. It starts with no features and keeps adding features one by one until the termination criterion is met. The SFS-QSAR tool uses the Mlxtend module (available at https://rasbt.github.io/mlxtend/ ) for feature selection process. To prepare the data, variables with minimal variance were removed using a 0.0001 variance threshold where multicollinearity was managed with a correlation threshold of 0.99. To evaluate the predictive nature of the developed models, four scoring functions were deployed including the likes of R 2 (determination coefficient) which measures how much variance in the dependent parameter is explained by the model, NMAE (normalized mean absolute error) which measures the average of the absolute errors between the predicted value and actual values, NMPD (normalized mean Poisson deviance) which measures errors for count-based data assuming a Poisson distribution and finally NMGD (normalized mean gamma deviance) which measures error for positive continuous data assuming gamma distribution. Eight (8) different models with SFS-QSAR were generated, four each for 5-fold and 0-fold cross validation. Unlike SFS, Genetic algorithm (GA) is a probabilistic optimization approach, which is inspired by the Darwinian evolution, employing processes analogous to inheritance, recombination, mutation and the survival of the fittest. Feature selection was performed through GeneticAlgorithm v.4.1_2 tool (available at http://teqip.jdvu.ac.in/QSAR_Tools/ ) [ 54 ]. As GA’s stochastic approach can create variability in the selected features, the model generation was repeated for 25 times. Similar to SFS, a variance threshold of 0.0001 was maintained, but the collinearity threshold was relaxed at 0.975. The algorithm was executed with default parameters; mutation rate of 0.3, 100 evolutionary cycles and crossover rate at 1. To eliminate the risk of overfitting and retain the relevant features only, the maximum descriptor count was limited to 7 for both datasets across all feature selection runs. Quality of the generated 2D-QSAR models were determined on the basis of Q 2 LOO and R 2 Pred ; which estimates the internal and external predictivity of the models. The details of these parameters have been provided in the Supplementary information. 2.2.3 Applicability domain (AD) of the models The AD is a theoretically determined descriptor space within which reliable and accurate predictions may be ascertained. The AD can help in estimating uncertainty, ensuring reliable predictions and also monitoring performance of the model. AD of the generated 2D-QSAR model was determined by Williams plot as well as by standardization approach. The Williams plot compares leverage values with standardized coefficients, helping to detect both response and structural outliers. It envisages the impact of each data point’s leverage against standardized residuals. When a data point’s leverage value exceeds the threshold hat value ( h* ), calculated as 3*p/n ( p : the number of descriptors + 1, and n : the number of data points in the training set), this datapoint is flagged as structural outlier. Response outliers are identified when any data point’s standardized residual surpasses ± 2.5. On the other hand, standardization approach is a relatively novel and more straightforward strategy for determining AD of QSAR models using basic theory of standardization[ 55 , 56 ]. 2.3 MolSHAP-based analysis In the next phase of this investigation, a relatively novel and unexplored tool was deployed to understand the contributions of different fragments within a chemical structure for exerting binding potential. The MolSHAP is a relatively new tool, recently developed by Tian et al, to determine the contributions of diverse fragments of a compound. MolSHAP is a model-agonistic framework which interprets the quantitative structure activity relationship (QSAR) by focusing on R -group contributions within a molecule rather than fingerprints or SMILES as a whole [ 57 ]. MolSHAP works best with series of compounds with a common core which is then extracted by the decomposition of the structure. The decomposition function of the tool breaks the structures down into a core, common for all the compounds in the series, and its R -groups at different positions. Once the compounds are decomposed, the next step involves R -group ranking where the contribution of each of the fragment is determined. The idea behind this step is to identify the fragments which contribute the most in exerting biological activity. For this, initially, two compounds are compared where only one R -group is different and the difference or similarity of the biological activity helps in evaluating the contribution of the fragment in that structure. This leads to ranking of the R -groups from the most important to least important to exhibit biological activity. The next stage in MolSHAP analysis involves training the compounds with different machine learning algorithms. In the current work, Gaussian Process Regressor was used as it was a part of MolSHAP software provided by Tian et al ( https://github.com/tiantz17/MolSHAP ) [ 57 ]. The major competence of MolSHAP lies in its ability to interpret the contribution of the diverse fragments in a structure while using a modified Shapley value-based interpretation. Unlike the traditional SHAP analysis, this tool does not remove the fragments from a structure, which may lead to the formation of an invalid or broken structure. On the contrary, the tool uses R -group masking task where the fragment in question is replaced by a background neutral fragment so as to retain the integrity of the structure. By systematically masking different R -groups and analyzing how predictions change, MolSHAP computes Shapley values that reflect each fragment’s contribution to activity. These interpretations are evaluated using the R -group ranking task, which involves comparing the R -group rankings predicted by Shapley values to the actual rankings determined from experimentally observed or model-predicted differences in compound activity. MolSHAP also provides the option of determining uncertainty estimation for models like Gaussian processes (GP) which helps in evaluating the models’ quality. Determination of the model quality helps in eliminating models with low confidence level while retaining the ones with better model quality. Finally, MolSHAP enables compound optimization by leveraging the calculated Shapley values. Starting from a lead compound, R -groups with low MolSHAP values are identified and substituted with fragments that have higher contributions. This guided R-group replacement strategy is iteratively applied to design new molecules that are predicted to have enhanced biological activity. The optimized molecules are screened and retained only if they show improved predicted potency, and their uncertainty is acceptably low. For this particular study, MolSHAP tool was used (available at https://github.com/tiantz17/MolSHAP ) where initially the dataset compounds were prepared by keeping three distinct column with ID, SMILES and Activity (response variable) as headers and subjected to decompose.py function of the module. This leads to generation of the core molecule and fragments of the dataset. Once the side chain decomposition was completed the molshap.py module of the tool was deployed to find the decomposition results, side chain contribution matrix, list of optimized compounds and a scattered plot of optimized compounds. 2.4 3D- QSAR modelling 2.4.1 Alignment of the dataset ligands In this current work, we used rigid body alignment of the structures for the 3D-QSAR, with the help of open-source application called MinAlignConf (available at https://github.com/ncordeirfcup/MinAlignConf ). 2.4.2 Model development The 3D-QSAR relationship analysis was performed using Open3DQSAR, an open-source framework for modelling steric and electrostatic molecular interactions [ 58 , 59 ]. The detailed methodology for the model development was reported previously [ 60 , 61 ] and a brief description is provided in the Supplementary information mainly highlighting the dataset division and feature selection methodologies as well as statistical parameters. Finally, contour maps were visualized at partial least squared coefficient threshold of + 0.005 (green) and − 0.005 (yellow) for steric fields; and + 0.003 (blue) and − 0.003 (red) for electrostatic fields [ 62 ]. 2.5 Molecular docking For molecular docking studies, the DNA-PK protein was initially curated from RCSB Protein Data Bank (PDB ID: 7OTW ) (available at https://www.rcsb.org/structure/7OTW ). Docking analysis was carried out using Autodock Vina [ 63 , 64 ]. The initial stage of molecular docking requires pre-processing of the protein structure which includes addition of all the missing hydrogen atoms, assigning Gasteiger partial atomic charges and subsequently converting the coordinate file into pdbqt format using AutoDock tools. Following the preparation of the protein structure, the 3D structures of the most active compound, the least active compound and the co-crystallized reference inhibitor AZD7648 were constructed. These ligands were then subjected to energy minimization and geometric optimization employing the MMFF94 force field to ensure accurate conformational stability prior to the docking simulations. The next step includes defining a grid box around the active site of the protein with dimensions of X = 157.625, Y = 174.369 and Z = 218.693 with extension of 30Å. The resulting docking poses and binding affinities were subsequently utilized for molecular dynamics (MD) simulation analysis. 2.6 MD Simulation analysis The protein-ligand complexes derived from the docking studies were utilized for MD simulation which was carried out with 100ns long run. The detailed methodology for the MD simulation study has been described in some of the previous investigations [ 51 , 65 ]. Brief description is however provided in the Supplementary information to describe the set-up of MD simulation and its analyses. 3. RESULTS AND DISCUSSION 3.1 2D-QSAR modelling As the strategy is outlined before, the 2D-QSAR models were developed with the intention of having better interpretable models without sacrificing the predictive quality of the models. To develop best linear model, two sets of alvaDesc descriptors sets were considered and these are (a) interpretable descriptors and (b) all descriptors [ 62 ] to check which set may provide the best solution. Naturally, the chance of getting a better QSAR model is always greater for the latter set since the former is just a subset of the latter. However, due to the fact the feature selection strategy is a complex method, it is always worthwhile to check which set produces the better result. If the second set is indeed found to generate the better results, it is required to check how much difference is found between the best models produced with by these sets. Evidently, a predictive model produced by the interpretable descriptors is capable of providing high statistical quality as well as mechanistic interpretability. Since, it is not wise to compromise the statistical results for sake of molecular interpretability, the current approach clarifies whether inclusion of all descriptors is indeed required. In this particular investigation, two distinct approaches were conceived to develop the 2D-QSAR models which required utilization of SFS and GA algorithms. The linear regression models were developed from the statistics resulted from the training and test sets. To consider the best predictive model from this investigation the average of Q 2 LOO ; an estimation of the internal predictivity, and R 2 Pred ; estimating the external predictivity was considered. The findings from the 2D-QSAR analysis are listed in Table 1 [ 51 ]. Table 1 Synopsis of the statistical parameters from 2D-QSAR models. Method Score Fold Interpretable descriptors All descriptors Q 2 LOO R 2 Pred Q 2 LOO R 2 Pred SFS R 2 0 0.642 ND 0.864 0.828 SFS R 2 5 0.695 ND 0.825 ND SFS NMAE 0 0.664 ND 0.764 ND SFS NMAE 5 0.689 ND 0.522 ND SFS NMPD 0 0.664 ND 0.833 ND SFS NMPD 5 0.685 ND 0.795 ND SFS NMGD 0 0.678 ND 0.814 ND SFS NMGD 5 0.683 ND 0.795 ND GA NA NA 0.762 0.208 0.827 ND #ND: Not Determined Since multiple models were generated here with different descriptor sets and model development strategy, we selected the most predictive model from maximum Q 2 LOO , which justified the internal predictivity of the model. Table 1 clearly demonstrates that the interpretable descriptors were inadequate to develop a predictive model as predictive as the models produced with all alvaDesc descriptors. The most predictive model was produced when the SFS-MLR model was constructed using R 2 as scoring function (without CV). This model (henceforth designated as Model 1) showed Q 2 LOO of 0.864 and R 2 Pred of 0.828. This SFS- MLR model is presented in Table 2 with 2D-QSAR equation and statistical parameters for the training and test sets. Overall, the model is generated with a satisfactory statistical predictivity and it is also evident from statistical parameters other than Q 2 LOO and R 2 Pred . Reports suggested that validation parameters based on rm 2 statistics are more robust than these two parameters [ 66 ]. The internal validation parameters rm 2 Train and the Δrm 2 Train values of the model were found as 0.814 and 0.059, respectively. At the same time, the rm 2 Test and ∆rm 2 Test values are 0.769 and 0.120, respectively. The observed vs. predicted pIC 50 values of 2D-QSAR model is depicted in Fig. 1. After that we checked the correlation matrix of this model (shown in Fig. 1) and it was observed that these descriptors are devoid of high inter-collinearity as the highest absolute correlation (Pearson r ) between any two descriptor is 0.45 (i.e., < 0.95). Y-randomization was performed by shuffling the pIC50 values of the compounds over 1000 iterations, yielding a c R 2 P value of 0.837. The relatively high score confirms the statistical reliability of the model and not generated by random chance. Table 2 Detailed statistical parameters of 2D-QSAR model. Equation Training set Test set pIC 50 =-6.404(± 0.844) * VE1sign_B(m) -2.456(± 0.595) * GATS8v + 3.866(± 0.729) * TDB03p + 1.815(± 0.28) * L3v − 0.333(± 0.066) * F02[C-N] + 0.257(± 0.051) * CATS3D_02_DL + 13.555(± 3.049) * WHALES20_IR + 5.46(± 1.756) N tr =59, R 2 = 0.900, N ts = 14, R 2 Pred /Q 2 F1 R 2 adj = 0.886, = 0.828, Q 2 F2 = 0.830, Q 2 LOO = 0.864, Q 2 F3 = 0.848, MAE = 0.029, RMSEP = 0.420, MSE = 0.341 rm 2 Test = 0.769, rm 2 Train = 0.814, ∆rm 2 Test = 0.120 ∆rm 2 Train = 0.059 Moreover, it is absolutely important to check the applicability domain (AD) for the establishment of the 2D-QSAR model. The Williams plot was generated for the Model 1, which is presented in the Fig. 2 . No structural or response outliers were found in this model. However, standardization approach detected one structural outlier in the training set and we retained this since no test set compound was identified as outlier. Figure 2 (B) provides an idea about the impact of each descriptor from the model M1. The model with the all descriptors shows insightful structural information of 6‑Anilino Imidazo[4,5‑c] pyridin-2-one derivatives. At the same time, out of 7 descriptors, two ( i.e., WHALES20_IR and VE1sign_B(m) ) showed high significance and three ( i.e., TDB03p , L3v and GATS8v ) demonstrated moderate significance. Two descriptors, namely CATS3D_02_DL and F02[C-N] depicted negligible significance. The descriptions of the descriptors of the model are outlined in Table 3 . Noticeably, most of the descriptors of the model is topological in nature and less interpretable. We first describe the WHALES20_IR descriptor which is found to be the most influential descriptor from Model 1 and it was found to have positive correlation towards the response variable. Weighted Holistic Atom Localization and Entity Shape ( WHALES ) descriptors are kind of 3D topological descriptors, which capture molecular shape and partial charges concurrently. Here IR of the WHALES20_IR denotes isolation and remoteness, and 20 is ratio percentile. Isolation depends on the local electron density and nearby significant group; remoteness is influenced by the topological distance in 3D structure. Molecules with different IR profiles may have different impact on the biological activity whereas 20 percentile value focuses on the less isolated regions, which is critical for the active site for the ligands. Overall, it can be concluded that the specific 3D shape of the compounds is important to determine the biological activity. The maximum value of this descriptor is found in 88 , which is also the most potent compound of the dataset with reported pIC 50 value of 8.674. On the other hand, another compound 29 that showed the lowest value of WHALES20_IR (i.e., -0.135) was found to possess an pIC 50 of 5.442. In Fig. 3 , we demonstrated and compared the structures of the most potent compound 88 and least potent compound 41 , which has WHALES20_IR value of -0.108. In the same figure, similar comparisons were performed with second and third most influential descriptors - VE1sign_B(m) and TDB03p . The second most important descriptor is VE1sign_B(m) , with the negative corelation with the response variable. It is a 2D matrix-based descriptors that is weighted with atomic mass of the compounds. The third most important descriptor is TDB03p , under 3D autocorrelations category. This depends on the polarizability of the compounds, as it is positively corelated with the response variable the higher polarizability enhances the binding potential of the compounds which is shown in Fig. 3 . Interestingly, the most potent compound 88 was found to have maximum value (i.e., 1.831) of this descriptor whereas the lowest descriptor value (i.e., 1.435) was found in a low active derivative 73 (pIC 50 = 5.714). Two least potent derivatives 20 and 41 had TDB03p values of 1.486 and 1.479, respectively. In comparison to 88 , all these low active compounds contained more bulky non-polar residues such as cyclopentyl, that may be responsible for low activity towards PKI enzyme. The fourth and the fifth most important descriptors are GATS8v and L3v respectively. Even though GATS8v belongs to the 2D autocorrelations category and L3v is a WHIM descriptor, both these are weighted by the van der Waals volume and both of them are found to be positively correlated with the response variable. Both descriptors bank on the van der Waals volume. Two descriptors with least significance are F02[C-N] and CATS3D_02_DL . The F02[C-N] belongs to 2D-atom pair descriptor which denotes the depicts the number of times the pair C and N appear within the molecular structure at topological distance of 2. Low value of this descriptor was found to be important. Finally, CATS descriptors are pharmacophore descriptors and CATS3D_02_DL quantifies the frequency of donor-lipophilic feature pairs at a defined 3D distance interval (bin 02). Positive correlation of this particular descriptor with the response variable signifies that the presence of this structural feature is beneficial to the binding potential of the compounds [ 67 ]. Detailed description of each molecular descriptor in the developed model has been provided in the Table S2 of Supplementary information. 3.2 MolSHAP analysis 2D-QSAR model provided satisfactory statistical results but provided us limited information about the contributions of various substitutions for increased or decreased potential towards DNA-PK. As it has been discussed earlier, MolSHAP is a comparatively novel strategy of QSAR model development that allows R -group analyses and may be considered as upgraded strategy of Free Wilson analysis. Here, the model generated with Gaussian Process Regressor (GPR) showed R 2 values of 0.987 and 0.844, respectively. However, it was necessary to more robust internal validation parameter and for this 5-fold cross-validation was performed with the training set that resulted in R 2 value of 0.583. Therefore, the model depicted satisfactory statistical predictivity. First, the contributions of the R groups were analysed and the results are depicted in Fig. 4 . In this figure the core common structure is shown and 10 most favourable and 10 most unfavourable fragments are provided. The R -group analyses revealed that the R 6 and R 7 positions (both attached to imidazo[4,5-c]- pyridine-2-one moiety) are the most important regions for determining increased biological activity. The presence of 4-methoxyphenyl moiety at position R 6 of the common structure appears as the most contributing fragment for increasing biological activity and it is followed by 4-hydroxyphenyl and 4-(2-methyl-2-phenylpropanenitrile) moieties. Noticeably, the latter fragments are found in two most potent compounds; 87 and 88 , with pIC 50 values of 8.357 and 8.674 respectively. Substitutions at positions R 1 , R 2 and R 4 also determine improved activity. Among this, the presence of methyl group at R 1 exhibits maximum impact with contribution of 0.83. Noticeably, the presence of isopropyl group at R 7 is responsible for lowering the biological potency. Noticeably, the presence of this moiety is found in 72 that has pIC 50 value 5.975. However, replacement of this isopropyl moiety with methyl group gives rise to 53 with significant improvement of biological activity (pIC 50 = 8.134). The MolSHAP analysis clearly indicates that the presence of methyl group at R 1 and R 4 is favourable whereas methyl group at R 5 is detrimental for increased potency. Similarly, influence of cyclopentyl was found to vary depending on its position. Substitution at R 6 was favourable whereas R 7 in unfavourable. Overall, the 2D-QSAR analysis highlights the favourable and unfavourable fragments but does not indicate the electrostatic and steric factors responsible for the increased or decreased potentials towards PKI enzyme. Therefore, we resorted to 3D-QSAR analyses to extract more information regarding these. 3.3 3D- QSAR modelling As seen in Table 3 , the better predictive quality was achieved through the deployment of FFD-SEL algorithm. The number of ligands considered to for training set was 59 while 14 datapoints were set aside for test set prediction. The Q 2 LOO (0.726), Q 2 LTO (0.760) and Q 2 LMO (0.736) values indicates that the model has strong internal predictivity, as well as external predictability ( R 2 Pred = 0.827). Additionally, the 3D-QSAR model accurately captures the role of both steric as well as the electrostatic fields in model development. Figure 5 shows a visual depiction of the contour maps. Table 3 Statistical parameters from 3D-QSAR analysis Parameter FFD-SEL UVE-PLS PC 4 4 N training 59 59 F-test 134.456 125.997 R 2 /SDEC 0.908/0.321 0.903/ 0.331 Q 2 LOO /SDEP 0.762/ 0.518 0.695/0.588 Q 2 LTO /SDEP 0.760/0.521 0.691/0.591 Q 2 LMO /SDEP 0.736/0.546 0.658/ 0.622 N test 14 14 R 2 Pred /SDEP 0.827/0.446 0.786/0.495 The contributions of the steric and electrostatic fields were found as 58% and 42%, respectively indicating the joint responsibility of both components in shaping the model. There is substantial consistency between the MolSHAP analyses and 3D-QSAR modelling. The presence of relatively bulky and electron rich 4- 2-methyl-2-phenylpropanenitrile moiety at R 6 -position was inserted into a steric favourable as well as electronegative fields. At the same position (i.e., R 6 ), the lowest active 41 contained relatively less electron rich cyclopentyl ring that was inserted into an electronegative field. Being a bulky group, the cyclopentyl was found close to neither steric favourable nor steric unfavourable fields. Compound 88 , the one with the highest binding affinity contains a bulky aromatic side chain that aligns well within the steric favourable region of the isocontour map. In addition, a methoxy and a methyl substituent are also positioned in sterically advantageous regions, prompting further enhancement in its binding potential. By contrast, compound 41 ; the one with the weakest binding affinity, lacks any bulky substituents near the steric favourable region, explaining one of the causes of its weak binding potential. Analysis of the electrostatic isocontour maps also revealed that compound 88 places its electron rich aromatic ring in closer proximity to an electrostatically favourable region, supporting beneficial interactions. On the other hand, compound 41 contains a hydrophobic cyclopentyl moiety within the same electronegative favourable region. Since non-polar groups are unable to interact effectively in such regions, this unfavourable placement likely results in decreased binding affinity. Overall, these observations provide important insights into the structure-activity relationship of DNA-dependent protein kinase inhibitors. 3.4 Molecular dynamics simulation The 2D-QSAR, MolSHAP and 3D-QSAR provided various information about the structural attributes for dataset compounds for enhanced potential regarding PKI. Nevertheless, it was necessary to understand the dynamic behaviours of the 6‑Anilino Imidazo[4,5‑c] pyridin-2-one’s derivatives with PKI. In order to understand this, we selected the most active compound 88 and one of the least active compounds 41 for MD simulation analyses. For molecular docking the protein structure provided by Hong et al was used in the current work. Notably, they also provided the docked complexes of Dactolisib and another dataset compound 78 . In the current work, we used the AZD7648-PKI complex as a reference to compare the dynamic behaviours of 88 and 41 . Interestingly, when the molecular docking was performed with 88 and 41 using Autodock Vina, the best docked pose of 88 replicated the structure of both AZD7648 as well as 78 (both provided by Hong et al). However, the docked conformation of 41 failed to replicate the similar conformation that led us to perform forced alignment of 41 with the docked pose of 78 to check the reason for this (Fig. S1 ). For this alignment, the . sdf form of 78 was used as a template whereas the structure of 41 was provided as SMILES. Subsequently, the aligned structure of 41 was minimized by OpenBabel tool using MMFF94 forcefield. The minimized structure of 41 was found to have multiple clashes at the binding site (shown in Fig. S1 ) that led us to proceed with the docked conformation of both 88 and 41 for 100ns MD simulations. After the MD simulation, we performed trajectory analyses, MM-GBSA analyses and per-residue decomposition analyses. The MM-GBSA results that are provided in Table 4 match the differences in experimental biological activities of 88 and 41 . Table 4 MM-GBSA binding energies (in kcal/mol) from MD simulations Complexes ΔE vdW ΔE elec ΔG polar ΔG nonpolar TΔS ΔG bind(T) pIC 50 88 -54.58 -7.08 19.86 -6.37 -22.57 -25.6 8.674 41 -44.61 -6.32 16.42 -5.23 -29.64 -10.1 5.000 AZD7648 -47.66 -7.13 17.71 -4.95 -18.89 -23.14 - More importantly, the best active 88 depicted significantly increased van der Waals interactions and considerably higher hydrogen bond interactions than low active 41 . Even though the nonpolar solvation energy of 88 was found to be more favourable, the polar solvation energy of the compound was less favourable for 88 that may be due to its larger molecular weight. The trajectory analyses, performed with 88 , 41 and AZD7684 revealed sufficient dynamic stabilities of all these ligands during MD simulations. However, slightly increased fluctuations were noted for low active 41 (average RMSF = 1.66 Å) as compared to 88 (average RMSF = 1.55 Å). Furthermore, as shown in Fig. 6 , 88 demonstrated higher number of hydrogen bond interactions in comparison with 41 further confirming that the former may produce increased electrostatic interactions with the receptor. In Fig. 7 , the last poses of these two compounds are depicted. However, to understand more specifically about the interactions between these ligands and DNA-PK, we performed the per residue decomposition analyses with the amino acids located within 10Å radius of the ligand. Noticeably, when we compare the per residue decomposition analyses results of 88 and 41 , the former depicted increased interactions with the receptor. In one hand, the electrostatic interactions of 88 with Lys3753 and Leu3806 were found to be significantly greater than 41 whereas van der Waals interaction of 88 with Trp3806 was found to be predominant in determining its higher binding potential. In this regard, we analysed the hydrogen bonds formed between these two ligands and the receptor. The trajectory analyses also revealed that the hydrogen bond interactions between pyridine nitrogen of 88 and Leu3806 may play the most significant role. Apart from that, the oxygen atom of the methoxy group of 88 forms hydrogen bond interactions with Lys3753. Compound 41 fails to show such significant interactions with the receptor and in this case, it is the Ser3731 that forms maximum hydrogen bond interaction with this compound. Noticeably, some π- π interactions were found to play pivotal role in determining the binding affinity. Interestingly, trajectory analyses revealed that almost 97% frames were identified in the last 50 ns MD simulation run where π- π interactions were found between heteroaromatic residues of 88 and Trp3805. However, only negligible number of frames were found where the aromatic rings of 41 engaged in π- π interactions with Trp3805 and Tyr3791. All these clearly indicate that it is the overall 3D topology of 41 that may hinder it to mimic the binding conformation adopted by 88 and other higher active compounds and it ultimately leads to lower activity towards DNA-PK. 3.5 MolSHAP based novel compound design Previously, we discussed development predictive MolSHAP QSAR modelling to understand favourable and unfavourable fragments. However, the method is also used to design and propose new compounds. Here, total 239 compounds were optimized and ranked as per the predicted pIC 50 values. We picked six structures with predicted pIC 50 greater than 88 (the most active compound of the dataset) and their structures were found that are depicted in Fig. 8 . However, these MolSHAP predicted structures were screened with the 2D-QSAR model, which was so far proved to be the most predictive model, MS02 and MS06 were found as structural outliers by the standardization method. Significantly, the molecular docking performed with these six compounds also depicted that except these structural outliers, the best poses of remaining four MolSHAP predicted compounds superimposed well with the bound conformation of AZD or docked pose of 78 (Provided by Hong et al). Now, after screening with our 2D-QSAR model, these four compounds exhibited predicted pIC 50 values as follows: MS01: 8.593, MS03: 8.397, MS04: 7.686 and MS05: 8.079. Therefore, as per the predictions provided by both MolSHAP and 2D-QSAR models, MS01 emerged as the most promising structure designed by MolSHAP. In order to ensure further about its reliability, we performed 100ns MD simulations with the docked conformation of it. This compound depicted ΔG bind(T) of -20.43 kCal/mol, which is very close to AZD7684. The energy contributions are as follows: ΔE vdW = -48.03 kCal/mol, ΔE elec = -8.47 kCal/mol, ΔG polar = 19.48 kCal/mol, ΔG nonpolar = -5.46 kCal/mol and TΔS = -22.04 kCal/mol. Noticeably, this compound has highly satisfactory polar interactions that is even greater than that observed for 88 . Additionally, trajectory analyses (shown in Supplementary information) depicted that MS01 maintains conformational stability and forms 5–7 hydrogen bond interactions during MD run. Analyses of the last pose of MS01 after MD simulation and the results of per residue decomposition analyses hint that MS01 may form stable hydrogen bond interactions with Leu3806 and Glu3804. Furthermore, hydrogen bond analyses indicated that its methoxy residue may also form stable hydrogen bond interactions with Asn3926. Similar to 88 , it also shows stable π-π interactions with Trp3805 as around 99% of frames in the last 50ns MD run depicted its existence and according to per residue decomposition analyses, van der Waals interactions with this amino acid residue may play predominant role in its binding with the receptor as shown in Fig. 9 . 4. CONCLUSION DNA-PK inhibitors have emerged as one of the leading candidates as radiosensitizers and the Hong et al designed a series of 6‑Anilino Imidazo[4,5‑c]pyridin-2-one derivatives through the much-coveted scaffold hopping strategy. This particular investigation, seeking to unravel the structural requirements of the DNA-PK inhibitors, constitutes of performing 2D-QSAR and 3D-QSAR analysis. Apart from the conventional QSAR analyses, the study also focuses on understanding the contributions of different structural fragments of the dataset ligands. This is followed by executing molecular dynamics simulation analysis for ligands with greatest and lowest biological activities. Finally, with the help of MolSHAP, new ligands were designed and the ligand with the best predicted activity was subjected to MD simulation analysis to reveal the various molecular interactions while binding with the receptor protein. Firstly, the 2D-QSAR analysis retrieved satisfactory results with both internal and external predictivities of 0.886 and 0.828 respectively. Furthermore, the model also brought forward the most significant descriptors from the dataset. It was found out that topological descriptor (WHALES20_IR) leverages a crucial influence on the binding potential of the ligands. The 3D topology of the structures provides an important idea about the electron density of the various atoms in the structure which is relevant in exerting binding capabilities of the ligands. Subsequently, the presence of bulky steric groups may influence negatively to the pIC 50 of the ligands. Descriptors from the 2D-QSAR analyses strongly indicate the importance of partial charges and van der Waals volume, exerting key influence on the binding potential. This was further supported by the MolSHAP analysis which clearly states the contributions of different fragments of the compounds in the dataset. From the MolSHAP analysis, it was found out that 4-methoxyphenyl moiety had the highest positive contribution in showing greater binding potential. The greater electron density of the group holds a significant importance that is also backed up by the findings from the 2D-QSAR analyses. Similar effects can be observed for the hydroxyphenyl moiety which is a strong electron donating group, having higher density of electrons in the structure. This phenomenon was also observed with 2D-QSAR analyses which heavily relied on the contributions of 3D topology and presence of electron rich regions of the dataset ligands. The findings from the 3D-QSAR analyses were found to be consistent with the information extracted from MolSHAP analysis which indicated the significance of electron rich 4-methoxyphenyl group in shaping the binding potential of the ligands. The 3D-QSAR isocontour maps also provides evidence of the electronegative fragment of the compound 88 being inserted into the electrostatic favourable region. The significance of the fragment was further consolidated by the fact that due to its bulky nature, it was also present in the steric favourable region of the isocontour map. On the other hand, for compounds with lesser binding affinity, the less electronegative fragment like cyclopentyl moiety was in close proximity to the electronegative region of the isocontour map. The consistency between the findings from the 2D, 3D and MolSHAP analyses encouraged us into looking at findings from the MD simulations. To understand the interactions between the ligands and receptor, 100ns MD simulation analyses were employed which gave a clear indication of the participating interactive force between them. The MM-GBSA analysis for both the highest and lowest active compounds along with the bound ligand AZD7684 provided insights on how electrostatic interactions might play an important role in determining the binding potential of the ligands. The trajectory analysis revealed that compound 88 participated in more hydrogen bond interactions than 41 which leads to its greater binding potential. The per-residue decomposition analysis provided evidence of electrostatic interactions between 88 and amino acid residues Lys3753 and Leu3806; while the van der Waals interactions with Trp3806 was the predominant factor contributing to the binding affinity of the ligand. These interactions were either absent or very weak in case of compound 41 ; leading to its lower biological potential. Finally, it was revealed that π-π interaction between the heterocyclic structure of 88 and Trp3805 was captured in almost 97% frames of the last 50ns of the MD run which was scarce for compound 41 . Upon understanding the ligand-receptor interaction, the novel designed molecule through MolSHAP was subjected to MD simulation analysis. The best predictive compound MS01 was found to exert similar MM-GBSA analysis to that of AZD7684 and the polar interactions were found to be more potent than that of 88 . The per-residue analysis of the ligand and receptor revealed stable H-bond interactions similar to 88 . Furthermore, MS01 also exhibited π-π interactions with Trp3805 which was found to be a crucial factor in exerting binding potential of the compounds. The findings from this investigation can be of enormous help in designing new DNA dependent protein kinase inhibitors as the structural requirements along with the relevant interactions with the receptor amino acids have been explored in details. Furthermore, the entire work is based on open-access tools that require no subscription cost and for this, the models are reproducible. Declarations CRediT authorship contribution statement S. Mitra: Writing – original draft, Investigation, Formal analysis, Validation, Resources, Data curation, Software. R. K. Dolai: Investigation, Data curation, Writing – original draft, Software. N. Ghosh: Writing – review & editing, Formal analysis, Conceptualisation, Methodology, Funding acquisition, Investigation, Supervision. S. C. Mandal: Formal analysis, Validation, Resources. A. K. Halder: Writing – review & editing, Conceptualisation, Visualisation, Validation, Supervision, Investigation, Formal analysis, Software. Funding This current work was funded by the Department of Science and Technology and Biotechnology, Govt. of West Bengal, India Vide Memo. 2027 (Sanc.)/STBT-11012 (19)/ 6/2023-ST SEC, dated 24–01-2024. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data Availability Data is provided within the manuscript or supplementary information files References Delaney G, Jacob S, Featherstone C, Barton M (2005) The role of radiotherapy in cancer treatment: estimating optimal utilization from a review of evidence-based clinical guidelines. Cancer 104:1129-37. https://doi.org/10.1002/cncr.21324. Moding EJ, Kastan MB, Kirsch DG (2013) Strategies for optimizing the response of cancer and normal tissues to radiation. Nat Rev Drug Discov 12:526-42. https://doi.org/10.1038/nrd4003. Pignon JP, Bourhis J, Domenge C, Designe L (2000) Chemotherapy added to locoregional treatment for head and neck squamous-cell carcinoma: three meta-analyses of updated individual data. MACH-NC Collaborative Group. Meta-Analysis of Chemotherapy on Head and Neck Cancer. Lancet 355:949-55. Bentzen SM, Trotti A (2007) Evaluation of early and late toxicities in chemoradiation trials. J Clin Oncol 25:4096-103. https://doi.org/10.1200/JCO.2007.13.3983. Ewing D (1998) The Oxygen Fixation Hypothesis: A Reevaluation. 21:355-61. Jing X, Yang F, Shao C, Wei K, Xie M, Shen H, et al. (2019) Role of hypoxia in cancer therapy by regulating the tumor microenvironment. Mol Cancer 18:157. https://doi.org/10.1186/s12943-019-1089-9. Wilson WR, Hay MP (2011) Targeting hypoxia in cancer therapy. Nat Rev Cancer 11:393-410. https://doi.org/10.1038/nrc3064. Overgaard J (2007) Hypoxic radiosensitization: adored and ignored. J Clin Oncol 25:4066-74. https://doi.org/10.1200/JCO.2007.12.7878. Liew LP, Shome A, Wong WW, Hong CR, Hicks KO, Jamieson SMF, et al. (2023) Design, Synthesis and Anticancer Evaluation of Nitroimidazole Radiosensitisers. Molecules 28. https://doi.org/10.3390/molecules28114457. Groelly FJ, Fawkes M, Dagg RA, Blackford AN, Tarsounas M (2023) Targeting DNA damage response pathways in cancer. Nat Rev Cancer 23:78-94. https://doi.org/10.1038/s41568-022-00535-5. O'Connor MJ (2015) Targeting the DNA Damage Response in Cancer. Mol Cell 60:547-60. https://doi.org/10.1016/j.molcel.2015.10.040. Paques F, Haber JE (1999) Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae. Microbiol Mol Biol Rev 63:349-404. https://doi.org/10.1128/MMBR.63.2.349-404.1999. Mahaney BL, Meek K, Lees-Miller SP (2009) Repair of ionizing radiation-induced DNA double-strand breaks by non-homologous end-joining. Biochem J 417:639-50. https://doi.org/10.1042/BJ20080413. Scully R, Panday A, Elango R, Willis NA (2019) DNA double-strand break repair-pathway choice in somatic mammalian cells. Nat Rev Mol Cell Biol 20:698-714. https://doi.org/10.1038/s41580-019-0152-0. Shao Z, Davis AJ, Fattah KR, So S, Sun J, Lee KJ, et al. (2012) Persistently bound Ku at DNA ends attenuates DNA end resection and homologous recombination. DNA Repair (Amst) 11:310-6. https://doi.org/10.1016/j.dnarep.2011.12.007. Downs JA, Jackson SP (2004) A means to a DNA end: the many roles of Ku. Nat Rev Mol Cell Biol 5:367-78. https://doi.org/10.1038/nrm1367. Mimori T, Hardin JA (1986) Mechanism of interaction between Ku protein and DNA. J Biol Chem 261:10375-9. Hartley KO, Gell D, Smith GC, Zhang H, Divecha N, Connelly MA, et al. (1995) DNA-dependent protein kinase catalytic subunit: a relative of phosphatidylinositol 3-kinase and the ataxia telangiectasia gene product. Cell 82:849-56. https://doi.org/10.1016/0092-8674(95)90482-4. Ding Q, Reddy YV, Wang W, Woods T, Douglas P, Ramsden DA, et al. (2003) Autophosphorylation of the catalytic subunit of the DNA-dependent protein kinase is required for efficient end processing during DNA double-strand break repair. Mol Cell Biol 23:5836-48. https://doi.org/10.1128/MCB.23.16.5836-5848.2003. Mahajan KN, Nick McElhinny SA, Mitchell BS, Ramsden DA (2002) Association of DNA polymerase mu (pol mu) with Ku and ligase IV: role for pol mu in end-joining double-strand break repair. Mol Cell Biol 22:5194-202. https://doi.org/10.1128/MCB.22.14.5194-5202.2002. Niewolik D, Pannicke U, Lu H, Ma Y, Wang LC, Kulesza P, et al. (2006) DNA-PKcs dependence of Artemis endonucleolytic activity, differences between hairpins and 5' or 3' overhangs. J Biol Chem 281:33900-9. https://doi.org/10.1074/jbc.M606023200. Niewolik D, Peter I, Butscher C, Schwarz K (2017) Autoinhibition of the Nuclease ARTEMIS Is Mediated by a Physical Interaction between Its Catalytic and C-terminal Domains. J Biol Chem 292:3351-65. https://doi.org/10.1074/jbc.M116.770461. Fulop GM, Phillips RA (1990) The scid mutation in mice causes a general defect in DNA repair. Nature 347:479-82. https://doi.org/10.1038/347479a0. Biedermann KA, Sun JR, Giaccia AJ, Tosto LM, Brown JM (1991) scid mutation in mice confers hypersensitivity to ionizing radiation and a deficiency in DNA double-strand break repair. Proc Natl Acad Sci U S A 88:1394-7. https://doi.org/10.1073/pnas.88.4.1394. Curtin NJ (2012) DNA repair dysregulation from cancer driver to therapeutic target. Nat Rev Cancer 12:801-17. https://doi.org/10.1038/nrc3399. Davidson D, Amrein L, Panasci L, Aloyz R (2013) Small Molecules, Inhibitors of DNA-PK, Targeting DNA Repair, and Beyond. Front Pharmacol 4:5. https://doi.org/10.3389/fphar.2013.00005. Lee TW, Wong WW, Dickson BD, Lipert B, Cheng GJ, Hunter FW, et al. (2019) Radiosensitization of head and neck squamous cell carcinoma lines by DNA-PK inhibitors is more effective than PARP-1 inhibition and is enhanced by SLFN11 and hypoxia. Int J Radiat Biol 95:1597-612. https://doi.org/10.1080/09553002.2019.1664787. Sibanda BL, Chirgadze DY, Blundell TL (2010) Crystal structure of DNA-PKcs reveals a large open-ring cradle comprised of HEAT repeats. Nature 463:118-21. https://doi.org/10.1038/nature08648. Sibanda BL, Chirgadze DY, Ascher DB, Blundell TL (2017) DNA-PKcs structure suggests an allosteric mechanism modulating DNA double-strand break repair. Science 355:520-4. https://doi.org/10.1126/science.aak9654. Yin X, Liu M, Tian Y, Wang J, Xu Y (2017) Cryo-EM structure of human DNA-PK holoenzyme. Cell Res 27:1341-50. https://doi.org/10.1038/cr.2017.110. Cao R, Zeng H, Zhang H (2009) 3D-QSAR studies on a series of inhibitors docked into a new homology model of the DNA-PK receptor. Curr Pharm Des 15:3796-825. https://doi.org/10.2174/138161209789649484. Ihmaid S, Al-Rawi J, Bradley C, Angove MJ, Robertson MN, Clark RL (2011) Synthesis, structural elucidation, DNA-PK inhibition, homology modelling and anti-platelet activity of morpholino-substituted-1,3-naphth-oxazines. Bioorg Med Chem 19:3983-94. https://doi.org/10.1016/j.bmc.2011.05.032. Clapham KM, Rennison T, Jones G, Craven F, Bardos J, Golding BT, et al. (2012) Potent enantioselective inhibition of DNA-dependent protein kinase (DNA-PK) by atropisomeric chromenone derivatives. Org Biomol Chem 10:6747-57. https://doi.org/10.1039/c2ob26035b. Maira SM, Stauffer F, Brueggen J, Furet P, Schnell C, Fritsch C, et al. (2008) Identification and characterization of NVP-BEZ235, a new orally available dual phosphatidylinositol 3-kinase/mammalian target of rapamycin inhibitor with potent in vivo antitumor activity. Mol Cancer Ther 7:1851-63. https://doi.org/10.1158/1535-7163.MCT-08-0017. Mukherjee B, Tomimatsu N, Amancherla K, Camacho CV, Pichamoorthy N, Burma S (2012) The dual PI3K/mTOR inhibitor NVP-BEZ235 is a potent inhibitor of ATM- and DNA-PKCs-mediated DNA damage responses. Neoplasia 14:34-43. https://doi.org/10.1593/neo.111512. Hardcastle IR, Cockcroft X, Curtin NJ, El-Murr MD, Leahy JJ, Stockley M, et al. (2005) Discovery of potent chromen-4-one inhibitors of the DNA-dependent protein kinase (DNA-PK) using a small-molecule library approach. J Med Chem 48:7829-46. https://doi.org/10.1021/jm050444b. Zhao Y, Thomas HD, Batey MA, Cowell IG, Richardson CJ, Griffin RJ, et al. (2006) Preclinical evaluation of a potent novel DNA-dependent protein kinase inhibitor NU7441. Cancer Res 66:5354-62. https://doi.org/10.1158/0008-5472.CAN-05-4275. Cano C, Saravanan K, Bailey C, Bardos J, Curtin NJ, Frigerio M, et al. (2013) 1-substituted (Dibenzo[b,d]thiophen-4-yl)-2-morpholino-4H-chromen-4-ones endowed with dual DNA-PK/PI3-K inhibitory activity. J Med Chem 56:6386-401. https://doi.org/10.1021/jm400915j. Munck JM, Batey MA, Zhao Y, Jenkins H, Richardson CJ, Cano C, et al. (2012) Chemosensitization of cancer cells by KU-0060648, a dual inhibitor of DNA-PK and PI-3K. Mol Cancer Ther 11:1789-98. https://doi.org/10.1158/1535-7163.MCT-11-0535. Shinohara ET, Geng L, Tan J, Chen H, Shir Y, Edwards E, et al. (2005) DNA-dependent protein kinase is a molecular target for the development of noncytotoxic radiation-sensitizing drugs. Cancer Res 65:4987-92. https://doi.org/10.1158/0008-5472.CAN-04-4250. Fok JHL, Ramos-Montoya A, Vazquez-Chantada M, Wijnhoven PWG, Follia V, James N, et al. (2019) AZD7648 is a potent and selective DNA-PK inhibitor that enhances radiation, chemotherapy and olaparib activity. Nat Commun 10:5065. https://doi.org/10.1038/s41467-019-12836-9. Goldberg FW, Finlay MRV, Ting AKT, Beattie D, Lamont GM, Fallan C, et al. (2020) The Discovery of 7-Methyl-2-[(7-methyl[1,2,4]triazolo[1,5-a]pyridin-6-yl)amino]-9-(tetrahydro-2H-pyran-4-yl)-7,9-dihydro-8H-purin-8-one (AZD7648), a Potent and Selective DNA-Dependent Protein Kinase (DNA-PK) Inhibitor. J Med Chem 63:3461-71. https://doi.org/10.1021/acs.jmedchem.9b01684. Willoughby CE, Jiang Y, Thomas HD, Willmore E, Kyle S, Wittner A, et al. (2020) Selective DNA-PKcs inhibition extends the therapeutic index of localized radiotherapy and chemotherapy. J Clin Invest 130:258-71. https://doi.org/10.1172/JCI127483. Zenke FT, Zimmermann A, Sirrenberg C, Dahmen H, Kirkin V, Pehl U, et al. (2020) Pharmacologic Inhibitor of DNA-PK, M3814, Potentiates Radiotherapy and Regresses Human Tumors in Mouse Models. Mol Cancer Ther 19:1091-101. https://doi.org/10.1158/1535-7163.MCT-19-0734. Chaplin AK, Hardwick SW, Liang S, Kefala Stavridi A, Hnizda A, Cooper LR, et al. (2021) Dimers of DNA-PK create a stage for DNA double-strand break repair. Nat Struct Mol Biol 28:13-9. https://doi.org/10.1038/s41594-020-00517-x. Liang S, Blundell TL (2023) Human DNA-dependent protein kinase activation mechanism. Nat Struct Mol Biol 30:140-7. https://doi.org/10.1038/s41594-022-00881-w. Liang S, Thomas SE, Chaplin AK, Hardwick SW, Chirgadze DY, Blundell TL (2022) Structural insights into inhibitor regulation of the DNA repair protein DNA-PKcs. Nature 601:643-8. https://doi.org/10.1038/s41586-021-04274-9. Hong CR, Liew LP, Wong WW, Dickson BD, Cheng G, Shome A, et al. (2024) Identification of 6-Anilino Imidazo[4,5-c]pyridin-2-ones as Selective DNA-Dependent Protein Kinase Inhibitors and Their Application as Radiosensitizers. J Med Chem 67:12366-85. https://doi.org/10.1021/acs.jmedchem.4c01120. Sushko I, Novotarskyi S, Korner R, Pandey AK, Rupp M, Teetz W, et al. (2011) Online chemical modeling environment (OCHEM): web platform for data storage, model development and publishing of chemical information. J Comput Aided Mol Des 25:533-54. https://doi.org/10.1007/s10822-011-9440-2. Mauri A. alvaDesc: A Tool to Calculate and Analyze Molecular Descriptors and Fingerprints. In: Roy K, editor. Ecotoxicological QSARs. New York, NY: Springer US; 2020. p. 801-20. Mitra S, Halder AK, Koley A, Ghosh N, Panda P, Mandal SC, et al. (2024) Unveiling structural determinants for FXR antagonism in 1,3,4-trisubstituted-Pyrazol amide derivatives: A multi-scale in silico modelling approach. Comput Biol Med 180:108991. https://doi.org/10.1016/j.compbiomed.2024.108991. Danishuddin, Khan AU (2016) Descriptors and their selection methods in QSAR analysis: paradigm for drug design. Drug Discov Today 21:1291-302. https://doi.org/10.1016/j.drudis.2016.06.013. Lesani S, Tavalla M, Eslami G, Boozhmehrani MJ (2025) Quantitative Structure-Activity Relationship Modeling and Molecular Docking Studies of TgCDPK1 Inhibitors in Toxoplasma gondii. Microbiologyopen 14:e70039. https://doi.org/10.1002/mbo3.70039. Ambure P, Aher RB, Gajewicz A, Puzyn T, Roy K (2015) “NanoBRIDGES” software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometrics and Intelligent Laboratory Systems 147:1-13. https://doi.org/https://doi.org/10.1016/j.chemolab.2015.07.007. Gramatica P (2007) Principles of QSAR models validation: internal and external. 26:694-701. https://doi.org/https://doi.org/10.1002/qsar.200610151. Roy K, Kar S, Ambure P (2015) On a simple approach for determining applicability domain of QSAR models. Chemometrics and Intelligent Laboratory Systems 145:22-9. https://doi.org/https://doi.org/10.1016/j.chemolab.2015.04.013. Tian T, Li S, Fang M, Zhao D, Zeng J (2024) MolSHAP: Interpreting Quantitative Structure-Activity Relationships Using Shapley Values of R-Groups. J Chem Inf Model 64:2236-49. https://doi.org/10.1021/acs.jcim.3c00465. Tosco P, Balle T (2011) Open3DQSAR: a new open-source software aimed at high-throughput chemometric analysis of molecular interaction fields. J Mol Model 17:201-8. https://doi.org/10.1007/s00894-010-0684-x. Tosco P, Balle T, Shiri F (2011) Open3DALIGN: an open-source software aimed at unsupervised ligand alignment. J Comput Aided Mol Des 25:777-83. https://doi.org/10.1007/s10822-011-9462-9. Mondal I, Halder AK, Pattanayak N, Mandal SK, Cordeiro M (2024) Shaping the Future of Obesity Treatment: In Silico Multi-Modeling of IP6K1 Inhibitors for Obesity and Metabolic Dysfunction. Pharmaceuticals (Basel) 17. https://doi.org/10.3390/ph17020263. Mitra S, Chatterjee S, Bose S, Panda P, Basak S, Ghosh N, et al. (2024) Finding structural requirements of structurally diverse alpha-glucosidase and alpha-amylase inhibitors through validated and predictive 2D-QSAR and 3D-QSAR analyses. J Mol Graph Model 126:108640. https://doi.org/10.1016/j.jmgm.2023.108640. Ghosh A, Panda P, Halder AK, Cordeiro M (2022) In silico characterization of aryl benzoyl hydrazide derivatives as potential inhibitors of RdRp enzyme of H5N1 influenza virus. Front Pharmacol 13:1004255. https://doi.org/10.3389/fphar.2022.1004255. Eberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. J Chem Inf Model 61:3891-8. https://doi.org/10.1021/acs.jcim.1c00203. Trott O, Olson AJ (2010) AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31:455-61. https://doi.org/10.1002/jcc.21334. Mitra S, Halder AK, Ghosh N, Mandal SC, Cordeiro M (2023) Multi-model in silico characterization of 3-benzamidobenzoic acid derivatives as partial agonists of Farnesoid X receptor in the management of NAFLD. Comput Biol Med 157:106789. https://doi.org/10.1016/j.compbiomed.2023.106789. Roy K, Kar S (2014) The rm2 metrics and regression through origin approach: reliable and useful validation tools for predictive QSAR models (Commentary on 'Is regression through origin useful in external validation of QSAR models?'). Eur J Pharm Sci 62:111-4. https://doi.org/10.1016/j.ejps.2014.05.019. Todeschini R, Consonni V (2008) Handbook of molecular descriptors. John Wiley & Sons, Place. Additional Declarations No competing interests reported. Supplementary Files Supplementaryinformation.docx Cite Share Download PDF Status: Published Journal Publication published 22 Dec, 2025 Read the published version in Journal of Computer-Aided Molecular Design → Version 1 posted Editorial decision: Revision requested 23 Oct, 2025 Reviews received at journal 27 Sep, 2025 Reviewers agreed at journal 17 Sep, 2025 Reviewers invited by journal 14 Sep, 2025 Editor assigned by journal 09 Sep, 2025 Submission checks completed at journal 02 Sep, 2025 First submitted to journal 28 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7479073","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":516637157,"identity":"faeffc8e-0c5e-48bf-9ebd-51d4065388cf","order_by":0,"name":"Soumya Mitra","email":"","orcid":"","institution":"Dr. B. C. Roy College of Pharmacy \u0026 Allied Health Sciences","correspondingAuthor":false,"prefix":"","firstName":"Soumya","middleName":"","lastName":"Mitra","suffix":""},{"id":516637158,"identity":"b39b93fe-21d0-44cb-ab7a-2c282207892b","order_by":1,"name":"Rakesh Kumar Dolai","email":"","orcid":"","institution":"Dr. B. C. Roy College of Pharmacy \u0026 Allied Health Sciences","correspondingAuthor":false,"prefix":"","firstName":"Rakesh","middleName":"Kumar","lastName":"Dolai","suffix":""},{"id":516637159,"identity":"7bf03c3a-a91f-4fcf-b16c-807080e5ae89","order_by":2,"name":"Nilanjan Ghosh","email":"","orcid":"","institution":"Jadavpur University","correspondingAuthor":false,"prefix":"","firstName":"Nilanjan","middleName":"","lastName":"Ghosh","suffix":""},{"id":516637160,"identity":"b99c7c2e-0406-4b58-ae26-081b677709bb","order_by":3,"name":"Subhash C. Mandal","email":"","orcid":"","institution":"Jadavpur University","correspondingAuthor":false,"prefix":"","firstName":"Subhash","middleName":"C.","lastName":"Mandal","suffix":""},{"id":516637161,"identity":"ab924cd0-96df-4b01-ad22-2919b81263d7","order_by":4,"name":"Amit Kumar Halder","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5UlEQVRIiWNgGAWjYFACxgYQkmE4wMD4mFgtjQ1ALTxALczGxFsD1cImTZR6/mmH2x8w7rDj4Tve+6y6oOaevDkD87GPX/BokbidCHTYmWQeyTPHzW7POFZsuLOBLXm2DD5rwFramHkMbqSx3eZtSEgwOMBjzCyBR4c8REs9j8H9Z2zFRGkxgGg5DLSFjY0ZpoXxAx4thkAtMxLbjgP9ksYsPeNYguHOZrZkZnxekbud/uDDx7ZqOb7jxxg/F9QkyJuzNx9m/IFPDwgkoDgVaAUzDyEtKMAAiAnbMgpGwSgYBSMJAACHjEx4MUWwcwAAAABJRU5ErkJggg==","orcid":"","institution":"Dr. B. C. Roy College of Pharmacy \u0026 Allied Health Sciences","correspondingAuthor":true,"prefix":"","firstName":"Amit","middleName":"Kumar","lastName":"Halder","suffix":""}],"badges":[],"createdAt":"2025-08-28 10:08:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7479073/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7479073/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10822-025-00733-x","type":"published","date":"2025-12-22T15:57:15+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":91737596,"identity":"940f6549-8b89-4fd2-a3d6-6043aa80dd5a","added_by":"auto","created_at":"2025-09-19 18:05:31","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2588578,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscriptv3.docx","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/2a3c26ef82a3252aa8238125.docx"},{"id":91736070,"identity":"fb2d7ae2-b939-4e9e-af40-f26419d88033","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1250134,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/54eea2ec296aefe1e04c8f74.tif"},{"id":91737080,"identity":"4901ee68-74aa-4040-a416-ce5e44fbea85","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":952976,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/69cb7dc73c53b9ba808bea25.tif"},{"id":91736088,"identity":"89ba0b9d-ffe3-40df-8f7a-05788a241947","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":706192,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/03b79e9dba6cb414efe5627a.tif"},{"id":91737014,"identity":"c6c30228-7f19-4dc9-872c-c9d0d50f9642","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":868704,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/a636f8e2058d8f97ad627536.tif"},{"id":91736073,"identity":"b03d5e23-1009-4c65-bad9-7f952ee167cd","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2193126,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/e1ec13b7294534659f1f26d8.tif"},{"id":91736079,"identity":"f0d92da6-8507-419d-9282-1085f3bc861f","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2290972,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/54ae178d76503397868d5ca9.tif"},{"id":91736076,"identity":"c65346ed-68df-4b8b-b673-198f7146e746","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1969210,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/e6fc55c526c6e866571508cb.tif"},{"id":91737081,"identity":"e1c00633-0e6e-44b1-8092-3ee3b8de2ed8","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":755770,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/3b37acce6fd6c26517659216.tif"},{"id":91736075,"identity":"5056d4c1-ec3b-47ec-8167-8e29dc2d604a","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1758404,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/b4e8766d252c6bf5f1bc1a5a.tif"},{"id":91736074,"identity":"b221911e-a1fe-4163-8395-04722114bd4b","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"json","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6659,"visible":true,"origin":"","legend":"","description":"","filename":"5a388a0a182a4037b3249b493b26cda1.json","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/c281007faef51a20b0bc25a9.json"},{"id":91737475,"identity":"f80084a9-d549-4023-ac80-fc1fafb6c7a0","added_by":"auto","created_at":"2025-09-19 17:57:31","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":597504,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/ab9a75581437654303818e47.docx"},{"id":91736080,"identity":"a7a2dcdc-5baa-4063-9480-dd858f4e5fa1","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"xml","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":197252,"visible":true,"origin":"","legend":"","description":"","filename":"5a388a0a182a4037b3249b493b26cda11enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/a715393d51249fa641f3a0b3.xml"},{"id":91736084,"identity":"d2674444-3929-44e0-a0f7-c4dda14afbeb","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"tif","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1250134,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/0113b4cad0d701629cbe01f4.tif"},{"id":91737088,"identity":"f3ee9658-1685-41f3-848f-eac700ba1562","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":952976,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/f7d8fb4ef67d4647a8ffaeb0.tif"},{"id":91737023,"identity":"f8618952-4e74-4acc-a0ea-38873324ada8","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"tif","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":706192,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/0e3bbedbcf3a319f37c5c2c2.tif"},{"id":91737085,"identity":"253cc196-717b-415b-b650-f1e9f13f491e","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":868704,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/217c679462f7dea640b46032.tif"},{"id":91737021,"identity":"12ccc232-c3fd-4b2a-983f-dae1f7d07a3a","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"tif","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2193126,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/639b4d5a06de450de225646a.tif"},{"id":91737091,"identity":"b0d11944-a385-47fd-a422-69305a3680e4","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2290972,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/fae323b05f4d80dc34e5783f.tif"},{"id":91736110,"identity":"2729d180-0a68-48bc-b3c4-5e7fc0c65016","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"tif","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1969210,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/b8819aab8d494006dc684657.tif"},{"id":91737087,"identity":"86f3dc4a-b38d-4567-8ff1-dea1f5454e7c","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"tif","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":755770,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/a4712b4777fef201a73a6ceb.tif"},{"id":91737597,"identity":"16b99b03-3b40-4624-9a1b-1f66d83f4e84","added_by":"auto","created_at":"2025-09-19 18:05:31","extension":"tif","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1758404,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.tif","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/417e03e0f652e7377d2513dd.tif"},{"id":91736101,"identity":"89429810-95e3-4dce-bbe6-5e8c0dc41501","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":192032,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/a4ce6d0c8858a1fa0ff47da8.jpeg"},{"id":91736108,"identity":"86b2464f-bdb9-4bfe-8ac6-ac896ef4a5ff","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":181958,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/835e5f3cff7e154c41b6bb53.jpeg"},{"id":91737082,"identity":"1577d37b-a9d4-4495-8d1c-11f91566ab3f","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"jpeg","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":86910,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/36f2270dc29d6448d7144870.jpeg"},{"id":91737018,"identity":"c1b12ad7-81b1-4202-8f5c-39d181220747","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"jpeg","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":130934,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/cdf652293bde81ea64b526d0.jpeg"},{"id":91736095,"identity":"b1e102d8-70ff-4bdb-aff6-54b19ef776db","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":435828,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/8ae53c23ab3517efa9338a75.jpeg"},{"id":91737478,"identity":"876d7f69-de7c-4157-9732-122018e5a32f","added_by":"auto","created_at":"2025-09-19 17:57:32","extension":"jpeg","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":381322,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/6dd3c766cf756d85af0966e4.jpeg"},{"id":91736099,"identity":"68ba40ea-db47-42e0-a7cc-b83c3caeb8d5","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":426396,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/61140582a9c3db99129d1046.jpeg"},{"id":91736092,"identity":"51e888a3-5cc5-4206-94e3-2d70eff046fa","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100172,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/b93688f38cd52d83b2cff91f.jpeg"},{"id":91736104,"identity":"14d9999c-4886-4e44-b4f1-4b5dbadd98e2","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"jpeg","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":307706,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/85be350ec797db1703727c83.jpeg"},{"id":91736096,"identity":"104da802-96d2-4812-9c9d-26858096046e","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126160,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/d2929dcae8dafedc0543843b.png"},{"id":91736098,"identity":"c8f6b10b-ca22-4fda-98d2-f16d3b26a3ac","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":104648,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/84e1d19062d91665e3025552.png"},{"id":91736100,"identity":"60a092fb-9deb-47ac-b6ed-e6104879a483","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":165746,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/84e89dee180f93def8f27cb0.png"},{"id":91737030,"identity":"0c8cf155-64d0-4b98-a00f-c3269503a8e4","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":204293,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/0f393cc70aeb89a91c8b7da9.png"},{"id":91737025,"identity":"5ea9d8ee-59bc-439f-bffa-138343be2c2f","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":249593,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/36118af8951f029b4ca23640.png"},{"id":91737035,"identity":"5357c986-c358-4654-89a6-3f6df258b4a7","added_by":"auto","created_at":"2025-09-19 17:41:32","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":153279,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/f14f4aa4dbe2b6acef525357.png"},{"id":91736105,"identity":"d0e4ede2-d420-4f0c-8411-5a48ab141273","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":215069,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/3056267162d37ae4cd55dc1e.png"},{"id":91736102,"identity":"24ab8822-9b27-407c-8394-ecd59386d4c4","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":181359,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/dc93dd4fba24f1c3a4ef45cb.png"},{"id":91736117,"identity":"a010665f-3319-4687-9c6a-200dd7fb02b8","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"png","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":161905,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/a2d927af461a09689c843eea.png"},{"id":91737090,"identity":"8910a3f6-7bec-4671-8b07-da1e638740d0","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"png","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28342,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/413f682c1e886a37e4457f67.png"},{"id":91736103,"identity":"9abd95e6-98bf-4476-979f-eb4971e7a148","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25261,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/3e480520e2d0584157dd9046.png"},{"id":91736112,"identity":"3aa6e9b2-861e-4282-a9b6-b8ea91933ef4","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"png","order_by":42,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55554,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/58e548222a2a7c3f2956e99b.png"},{"id":91736107,"identity":"2a570816-53d8-465b-ab02-f5bf14e21fd9","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":43,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75255,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/dd6788f00ad0af37f6f1f716.png"},{"id":91736109,"identity":"e803c081-2b19-4432-9e2a-09181e3b6db4","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":44,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58473,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/083011d6441aca3502c15e07.png"},{"id":91737027,"identity":"e70bd616-7982-422a-9416-442f733e2a87","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":45,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38775,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/b27d3b2e1405d07ce6f8a3ed.png"},{"id":91736115,"identity":"2bd05979-3865-42f4-944c-b2a61b9fb568","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"png","order_by":46,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65395,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/1d23fba3623dddef44853ba6.png"},{"id":91737089,"identity":"a32ee7e9-20d0-4559-92bd-6589c678e2df","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"png","order_by":47,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":61581,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/1c7b31210185af99f43e7fb2.png"},{"id":91736120,"identity":"5955bbec-3104-409b-95f3-147ff10067a2","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"png","order_by":48,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40957,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/5fd96fe46dc285feb0d0edb7.png"},{"id":91737034,"identity":"a2852244-ba0e-4e8b-8cf7-febef1a0c852","added_by":"auto","created_at":"2025-09-19 17:41:32","extension":"xml","order_by":49,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":195825,"visible":true,"origin":"","legend":"","description":"","filename":"5a388a0a182a4037b3249b493b26cda11structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/623134d0a1ab230446d7b505.xml"},{"id":91736123,"identity":"73ca82c9-7d56-42b9-92c1-f492a96ea869","added_by":"auto","created_at":"2025-09-19 17:33:32","extension":"html","order_by":50,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":213047,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/ae8aa12ff0014522210d9b8c.html"},{"id":91736062,"identity":"48904749-a599-4493-bb8c-047986e56068","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":88717,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e(A) Observed vs predicted pIC\u003csub\u003e50\u003c/sub\u003e values and (B) correlation matrix of 2D-QSAR model.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/93017607f23cbddc627a8818.png"},{"id":91737008,"identity":"3f2f7dad-e046-4e3d-8727-1f5c2d26b477","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":69353,"visible":true,"origin":"","legend":"\u003cp\u003e(A) Williams plot for model M1 and (B) relative significance of the descriptors of model M1 with respect to standardized coefficient values.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/e1e4d11ff64e36e93dd90288.png"},{"id":91736063,"identity":"5f58a9e1-0fb3-4506-822f-750b13250cda","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":146285,"visible":true,"origin":"","legend":"\u003cp\u003eSignificance of\u003cstrong\u003e \u003c/strong\u003eWHALES20_IR, \u003cem\u003eVE1sign_B(m)\u003c/em\u003e and \u003cem\u003eTDB03p \u003c/em\u003edescriptors in the dataset ligands\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/ce9a25ffed3b6d8758431cbb.png"},{"id":91736064,"identity":"ebb040c5-e37a-4966-8d6d-1ca7a2c0673a","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":171818,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eR\u003c/em\u003e-group analyses from MolSHAP\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/e548b3cd265d4b85b9be395b.png"},{"id":91736067,"identity":"5bc3d271-0db5-4db8-bb8a-8908d9e49103","added_by":"auto","created_at":"2025-09-19 17:33:31","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":329230,"visible":true,"origin":"","legend":"\u003cp\u003eIsocontour maps from 3D-QSAR modelling analyses of compounds 88 and 41 with respect to steric and electrostatic regions\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/515905013f14342f3a84cc02.png"},{"id":91737079,"identity":"38419374-429e-4f9d-84af-f50f1b349759","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":137689,"visible":true,"origin":"","legend":"\u003cp\u003ePlots obtained from trajectory analyses (A) ligand RMSD and (B) number of hydrogen bonds formed between amino acids and ligands\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/1699d4c34b98ac590ded730d.png"},{"id":91737078,"identity":"62c871c8-f1d8-4dc0-ac43-108a330590ac","added_by":"auto","created_at":"2025-09-19 17:49:31","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":313368,"visible":true,"origin":"","legend":"\u003cp\u003e(A) The receptor-ligand interactions for the 88-complex (left) and the 41-complex (right) extracted from the last frame of the MD simulations. (B) Per-residue decomposition analyses of compounds 88 (left) and 41 (right)\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/634938e7ba6a43771369a9ee.png"},{"id":91737011,"identity":"a7b3e76f-96d9-42cc-b703-3cbcd393e251","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":177359,"visible":true,"origin":"","legend":"\u003cp\u003eNovel ligands designed by MolSHAP\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/e010432b851b348a573e3f6b.png"},{"id":91737016,"identity":"a9c2d981-355b-46ae-8259-52b7c8185865","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":153819,"visible":true,"origin":"","legend":"\u003cp\u003e(A) Receptor ligand interaction between MS01 and receptor; (B) Per-residue decomposition analyses of MS01\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/54a80e90a2034a9c4925d5e3.png"},{"id":99172360,"identity":"c542a503-119d-4168-8558-409b3ad5ec94","added_by":"auto","created_at":"2025-12-29 16:08:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2553996,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/51c4aeb0-6de9-4b84-beb0-174abda6125b.pdf"},{"id":91737009,"identity":"9d773a64-3b25-44ee-ac16-a6310a8d6ff4","added_by":"auto","created_at":"2025-09-19 17:41:31","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":597504,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-7479073/v1/7f4d92cf0c14640381ab3bfe.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-scale in-silico modelling to unveil structural requirements for DNA-PK inhibitors as radiosensitizers and MolSHAP based design of novel ligands","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cp\u003eCancer remains one of the most complicated diseases to appear during the course of history with no definite therapeutic strategies to succeed. With over 1.4\u0026nbsp;million new cases detected in India alone, therapies related to the management of cancer is highly dependent on radiotherapy. Radiotherapy (RT) has been one of the most commonly used approaches in the management of different types of cancers with around 52.6% of the total population of cancer patients undergoing RT at different stages of the disease [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. RT, in combination with chemotherapy, along with radiosensitizers has become the typical approach for treating head and neck small cell carcinoma [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. While the strategy is beneficial in cases with detection of cancer, this therapeutic approach also presents with high occurrences of toxicity towards the unaffected cells and tissue [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. To combat the adverse effects associated with radiotherapy, several investigations have been dedicated to enhance the efficacy of the RT with minimal exposure of the patients to the radiation. One such strategy involves the introduction of radiosensitizers which are chemical compounds capable of increasing the efficiency of the RT. Radiosensitizers are effectively compounds which make the tumor cells more susceptible towards RT. Radiotherapy deals with ionizing radiation which can damage the DNA in the cancerous cells in irreversible manner. RT produces free radical, especially hydroxyl radicals, which promote breaking of the DNA double strand. The oxygen then reacts with the hydroxyl free radical to give rise to peroxide radicals which causes a more permanent damage towards the DNA strands [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. In case of tumors, they are often seen in regions with low oxygen content leading to hypoxic tumours. Having scarcity of oxygen often leads to resistance towards RT which is caused by the inability to form peroxide free radicals. To overcome the RT associated resistance, investigators have resorted to developing compounds like nitroimidazole, which can act as substitute for oxygen in those low oxygen environments and cause irreversible DNA damage[\u003cspan additionalcitationids=\"CR7 CR8\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Another strategy to counter the RT resistance is by developing protein kinase inhibitors that primarily stops the hypoxic cells from recovering from the DNA damage. In recent years, more intensive investigations have been carried out which focussed on the pathways that lead to DNA damage repair through development of modern radiosensitizers [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eRT has been responsible for creating double strand breaks (DSB) while inducing cytotoxic effects on the cells. Resistance towards RT can be indicated by the ability of the DNA to repair itself; primarily by two pathways; which include non-homologous end joining (NHEJ) and homologous recombination (HR) [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. However, the NHEJ plays the predominant role as it increases radiosensitivity of cells by mutating the NHEJ genes [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Within the NHEJ pathway, the Ku heterodimers rapidly locate the DSB and attach itself to the DNA terminal to prevent any further nucleolytic damage [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Irrespective of the structural configuration of the DNA, the Ku dimers show high binding affinity to the damaged ends. Due to its ring-like shape (toroidal) it can easily attach itself to the DNA terminal and start recruiting DNA-PK catalytic subunits (DNA-PKcs) along with the nuclease Artemis. Artemis is responsible for trimming the damaged ends of the DNA and together with the DNA-PKcs, they form a DNA-PK holoenzyme. The association of the Ku with DNA-PKcs increases the stability of the Ku-DNA complex and promotes a conformational shift in DNA-PKcs which activates the kinase function through autophosphorylation at multiple regulatory sites. Once activated, DNA-PKcs induces phosphorylation at the C-terminal of the autoinhibitory segment of Artemis, enabling to perform the nuclease activity Following end processing, the DNA-PKcs recruit gap-filling polymerases from the Pol X family, specifically DNA polymerase λ and DNA polymerase \u0026micro;. These enzymes restore the missing nucleotides, following which the DNA ligase IV-XRCC4-XLF complex completes the ligation to facilitate DNA repair [\u003cspan additionalcitationids=\"CR16 CR17 CR18 CR19 CR20 CR21\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. This loss of DNA-PK functions severely sensitises the DSBs [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e], that makes DNA-PK as major drug target [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Recently it was confirmed that the DNA-PK offers the higher radio sensitization than inhibitors of PARP-1 (poly (ADP-ribose) polymerase-1) in HNSCC (Head and neck squamous cell carcinoma) cells [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. The limited resolution of DNA-PK structural data first made it difficult to find strong and specific DNA-PK inhibitors [\u003cspan additionalcitationids=\"CR29 CR30\" citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], and dependent on the homology model [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] based on the phosphoinositide 3-kinase (PI3K) enzymes. Because of this, initial inhibitors like dactosilib, NU7441, KU0060648, and IC87361 show lack of precise selectivity to DNA-PK, they affected not only DNA-PK but also PI3K and other related enzymes in the PIKK (PI3K-related kinase) family [\u003cspan additionalcitationids=\"CR35 CR36 CR37 CR38\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. This lack of selectivity led to unwanted adverse effect or toxicities in preclinical studies, limiting their therapeutic potency [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. In recent time, more selective inhibitors to DNA-PK such as AZD7684, NU5455, Pepostertib (M3814), and 31t are identified and most of these are characterized at higher resolution in structural studies [\u003cspan additionalcitationids=\"CR42 CR43 CR44 CR45 CR46\" citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. Two of them, Peposertib with RT as well as AZD7648 in combination with pegylated doxorubicin or Olaparib, have progressed to clinical trials. In the current work, we collected a data set containing 73 6‑Anilino Imidazo [4,5‑c] pyridin-2-one\u0026rsquo;s derivatives that were recently reported as novel DNA-PK inhibitors and performed a range of in silico modelling analyses to understand the structural requirements in order to ensure that more robust design strategy may be adopted with such information for obtaining DNA-PK inhibitors with greater potency [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e"},{"header":"2. MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Dataset curation and preparation\u003c/h2\u003e\u003cp\u003eIn search of potent DNA-PK inhibitors, Hong et al. designed a series of novel 6Anilino Imidazo[4,5c] pyridin-2-one derivatives by deploying scaffold hopping techniques on an established PI3K and PIKK inhibitor Dactolisib [48]. The group developed new inhibitors and evaluated their inhibitory potential against the DNA-PK enzyme using the HotSpot assay platform. The nanomolar IC\u003csub\u003e50\u003c/sub\u003e values reported for the 73 compounds were subsequently transformed into pIC\u003csub\u003e50\u003c/sub\u003e (= ‒log\u003csub\u003e10\u003c/sub\u003e(IC\u003csub\u003e50\u003c/sub\u003e/10\u003csup\u003e9\u003c/sup\u003e) values.\u003c/p\u003e\u003cp\u003eThe SMILES notations of the dataset compounds are collected from the article of Hong et al [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. The dataset was prepared by converting the curated SMILES into their canonical forms via RDKit (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.rdkit.org/\u003c/span\u003e\u003cspan address=\"https://www.rdkit.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and were converted into .sdf files through Discovery Studio Visualizer tool (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://discover.3ds.com/discovery-studio-visualizer-download\u003c/span\u003e\u003cspan address=\"https://discover.3ds.com/discovery-studio-visualizer-download\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). This was followed by incorporation of the biological activity in terms of pIC\u003csub\u003e50\u003c/sub\u003e as the response variables required to develop QSAR models. Once the dataset was ready (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e of the Supplementary information), the ligands of the dataset were subjected to structural standardization by the CDK standardizer. The standardization process entails adding explicit hydrogen atoms, performing aromatization, cleaning the structure in both 2D and 3D, neutralizing, and removing any salts.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 2D-QSAR modelling\u003c/h2\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1 \u003cem\u003eDescriptor calculation\u003c/em\u003e\u003c/h2\u003e\u003cp\u003eFor descriptor calculation, the open-access webserver OCHEM (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ochem.eu/login/show.do\u003c/span\u003e\u003cspan address=\"https://ochem.eu/login/show.do\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was utilized where 73 dataset ligands were fed to compute alvaDesc descriptors (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.alvascience.com/alvadesc-descriptors/\u003c/span\u003e\u003cspan address=\"https://www.alvascience.com/alvadesc-descriptors/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) in .\u003cem\u003esdf\u003c/em\u003e format [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. The molecular structures of the 6anilino imidazo[4,5c] pyridin-2-one derivatives were carried out by the RDKit module of the webserver. The descriptor calculation took place in two steps, wherein the first step included computation of all the alvaDesc descriptors and the next step included computing only those alvaDesc descriptors which have better interpretability. The details of the interpretable descriptors have been provided in one of our previous works [51].The descriptor files (.\u003cem\u003ecsv\u003c/em\u003e) were prepared by adding the identifier of dataset ligands and their respective pIC\u003csub\u003e50\u003c/sub\u003e values as response variables.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.2.2 \u003cem\u003eDataset division and model set-up\u003c/em\u003e\u003c/h2\u003e\u003cp\u003eAs both the datasets with all AlvaDesc descriptors and interpretable descriptors were developed, they were segregated into training and test (or prediction) sets by using SFS-QSAR-tool_v2 (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/ncordeirfcup/SFS-QSAR-tool_v2\u003c/span\u003e\u003cspan address=\"https://github.com/ncordeirfcup/SFS-QSAR-tool_v2\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]. The datasets were divided while deploying activity sorting 4, and the ratio between training and prediction set were kept at 4:1.\u003c/p\u003e\u003cp\u003eIn regression analysis, feature selection helps in identifying the relevant features during model development [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. Two distinct feature selection modules were utilized for setting up the 2D-QSAR models; (i) sequential forward selection and (ii) genetic algorithm. The first method SFS works by iteratively adding the most significant feature that improves model performance until a stopping criterion is met. On the other hand, GA is a heuristic search inspired by natural selection [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]. Sequential forward selection (SFS) is a non-stochastic greedy algorithm which acts by selecting the optimum features in a dataset. It starts with no features and keeps adding features one by one until the termination criterion is met. The SFS-QSAR tool uses the Mlxtend module (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://rasbt.github.io/mlxtend/\u003c/span\u003e\u003cspan address=\"https://rasbt.github.io/mlxtend/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for feature selection process. To prepare the data, variables with minimal variance were removed using a 0.0001 variance threshold where multicollinearity was managed with a correlation threshold of 0.99. To evaluate the predictive nature of the developed models, four scoring functions were deployed including the likes of \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e (determination coefficient) which measures how much variance in the dependent parameter is explained by the model, \u003cem\u003eNMAE\u003c/em\u003e (normalized mean absolute error) which measures the average of the absolute errors between the predicted value and actual values, \u003cem\u003eNMPD\u003c/em\u003e (normalized mean Poisson deviance) which measures errors for count-based data assuming a Poisson distribution and finally \u003cem\u003eNMGD\u003c/em\u003e (normalized mean gamma deviance) which measures error for positive continuous data assuming gamma distribution. Eight (8) different models with SFS-QSAR were generated, four each for 5-fold and 0-fold cross validation.\u003c/p\u003e\u003cp\u003eUnlike SFS, Genetic algorithm (GA) is a probabilistic optimization approach, which is inspired by the Darwinian evolution, employing processes analogous to inheritance, recombination, mutation and the survival of the fittest. Feature selection was performed through GeneticAlgorithm v.4.1_2 tool (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://teqip.jdvu.ac.in/QSAR_Tools/\u003c/span\u003e\u003cspan address=\"http://teqip.jdvu.ac.in/QSAR_Tools/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. As GA\u0026rsquo;s stochastic approach can create variability in the selected features, the model generation was repeated for 25 times. Similar to SFS, a variance threshold of 0.0001 was maintained, but the collinearity threshold was relaxed at 0.975. The algorithm was executed with default parameters; mutation rate of 0.3, 100 evolutionary cycles and crossover rate at 1. To eliminate the risk of overfitting and retain the relevant features only, the maximum descriptor count was limited to 7 for both datasets across all feature selection runs. Quality of the generated 2D-QSAR models were determined on the basis of \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e; which estimates the internal and external predictivity of the models. The details of these parameters have been provided in the Supplementary information.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.2.3 \u003cem\u003eApplicability domain (AD) of the models\u003c/em\u003e\u003c/h2\u003e\u003cp\u003eThe AD is a theoretically determined descriptor space within which reliable and accurate predictions may be ascertained. The AD can help in estimating uncertainty, ensuring reliable predictions and also monitoring performance of the model. AD of the generated 2D-QSAR model was determined by Williams plot as well as by standardization approach. The Williams plot compares leverage values with standardized coefficients, helping to detect both response and structural outliers. It envisages the impact of each data point\u0026rsquo;s leverage against standardized residuals. When a data point\u0026rsquo;s leverage value exceeds the threshold hat value (\u003cem\u003eh*\u003c/em\u003e), calculated as \u003cem\u003e3*p/n\u003c/em\u003e (\u003cem\u003ep\u003c/em\u003e: the number of descriptors\u0026thinsp;+\u0026thinsp;1, and \u003cem\u003en\u003c/em\u003e: the number of data points in the training set), this datapoint is flagged as structural outlier. Response outliers are identified when any data point\u0026rsquo;s standardized residual surpasses\u0026thinsp;\u0026plusmn;\u0026thinsp;2.5. On the other hand, standardization approach is a relatively novel and more straightforward strategy for determining AD of QSAR models using basic theory of standardization[\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.3 MolSHAP-based analysis\u003c/h2\u003e\u003cp\u003eIn the next phase of this investigation, a relatively novel and unexplored tool was deployed to understand the contributions of different fragments within a chemical structure for exerting binding potential. The MolSHAP is a relatively new tool, recently developed by Tian et al, to determine the contributions of diverse fragments of a compound. MolSHAP is a model-agonistic framework which interprets the quantitative structure activity relationship (QSAR) by focusing on \u003cem\u003eR\u003c/em\u003e-group contributions within a molecule rather than fingerprints or SMILES as a whole [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. MolSHAP works best with series of compounds with a common core which is then extracted by the decomposition of the structure. The decomposition function of the tool breaks the structures down into a core, common for all the compounds in the series, and its \u003cem\u003eR\u003c/em\u003e-groups at different positions.\u003c/p\u003e\u003cp\u003eOnce the compounds are decomposed, the next step involves \u003cem\u003eR\u003c/em\u003e-group ranking where the contribution of each of the fragment is determined. The idea behind this step is to identify the fragments which contribute the most in exerting biological activity. For this, initially, two compounds are compared where only one \u003cem\u003eR\u003c/em\u003e-group is different and the difference or similarity of the biological activity helps in evaluating the contribution of the fragment in that structure. This leads to ranking of the \u003cem\u003eR\u003c/em\u003e-groups from the most important to least important to exhibit biological activity. The next stage in MolSHAP analysis involves training the compounds with different machine learning algorithms. In the current work, Gaussian Process Regressor was used as it was a part of MolSHAP software provided by Tian et al (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tiantz17/MolSHAP\u003c/span\u003e\u003cspan address=\"https://github.com/tiantz17/MolSHAP\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. The major competence of MolSHAP lies in its ability to interpret the contribution of the diverse fragments in a structure while using a modified Shapley value-based interpretation. Unlike the traditional SHAP analysis, this tool does not remove the fragments from a structure, which may lead to the formation of an invalid or broken structure. On the contrary, the tool uses \u003cem\u003eR\u003c/em\u003e-group masking task where the fragment in question is replaced by a background neutral fragment so as to retain the integrity of the structure. By systematically masking different \u003cem\u003eR\u003c/em\u003e-groups and analyzing how predictions change, MolSHAP computes Shapley values that reflect each fragment\u0026rsquo;s contribution to activity. These interpretations are evaluated using the \u003cem\u003eR\u003c/em\u003e-group ranking task, which involves comparing the \u003cem\u003eR\u003c/em\u003e-group rankings predicted by Shapley values to the actual rankings determined from experimentally observed or model-predicted differences in compound activity.\u003c/p\u003e\u003cp\u003eMolSHAP also provides the option of determining uncertainty estimation for models like Gaussian processes (GP) which helps in evaluating the models\u0026rsquo; quality. Determination of the model quality helps in eliminating models with low confidence level while retaining the ones with better model quality. Finally, MolSHAP enables compound optimization by leveraging the calculated Shapley values. Starting from a lead compound, \u003cem\u003eR\u003c/em\u003e-groups with low MolSHAP values are identified and substituted with fragments that have higher contributions. This guided R-group replacement strategy is iteratively applied to design new molecules that are predicted to have enhanced biological activity. The optimized molecules are screened and retained only if they show improved predicted potency, and their uncertainty is acceptably low.\u003c/p\u003e\u003cp\u003eFor this particular study, MolSHAP tool was used (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tiantz17/MolSHAP\u003c/span\u003e\u003cspan address=\"https://github.com/tiantz17/MolSHAP\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) where initially the dataset compounds were prepared by keeping three distinct column with ID, SMILES and Activity (response variable) as headers and subjected to \u003cem\u003edecompose.py\u003c/em\u003e function of the module. This leads to generation of the core molecule and fragments of the dataset. Once the side chain decomposition was completed the \u003cem\u003emolshap.py\u003c/em\u003e module of the tool was deployed to find the decomposition results, side chain contribution matrix, list of optimized compounds and a scattered plot of optimized compounds.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.4 3D- QSAR modelling\u003c/h2\u003e\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\u003ch2\u003e2.4.1 Alignment of the dataset ligands\u003c/h2\u003e\u003cp\u003eIn this current work, we used rigid body alignment of the structures for the 3D-QSAR, with the help of open-source application called MinAlignConf (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/ncordeirfcup/MinAlignConf\u003c/span\u003e\u003cspan address=\"https://github.com/ncordeirfcup/MinAlignConf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cem\u003e).\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\u003ch2\u003e2.4.2 Model development\u003c/h2\u003e\u003cp\u003eThe 3D-QSAR relationship analysis was performed using Open3DQSAR, an open-source framework for modelling steric and electrostatic molecular interactions [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e]. The detailed methodology for the model development was reported previously [\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e] and a brief description is provided in the Supplementary information mainly highlighting the dataset division and feature selection methodologies as well as statistical parameters. Finally, contour maps were visualized at partial least squared coefficient threshold of +\u0026thinsp;0.005 (green) and \u0026minus;\u0026thinsp;0.005 (yellow) for steric fields; and +\u0026thinsp;0.003 (blue) and \u0026minus;\u0026thinsp;0.003 (red) for electrostatic fields [\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Molecular docking\u003c/h2\u003e\u003cp\u003eFor molecular docking studies, the DNA-PK protein was initially curated from RCSB Protein Data Bank (PDB ID: \u003cem\u003e7OTW\u003c/em\u003e) (available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.rcsb.org/structure/7OTW\u003c/span\u003e\u003cspan address=\"https://www.rcsb.org/structure/7OTW\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Docking analysis was carried out using Autodock Vina [\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e]. The initial stage of molecular docking requires pre-processing of the protein structure which includes addition of all the missing hydrogen atoms, assigning Gasteiger partial atomic charges and subsequently converting the coordinate file into \u003cem\u003epdbqt\u003c/em\u003e format using AutoDock tools. Following the preparation of the protein structure, the 3D structures of the most active compound, the least active compound and the co-crystallized reference inhibitor AZD7648 were constructed. These ligands were then subjected to energy minimization and geometric optimization employing the MMFF94 force field to ensure accurate conformational stability prior to the docking simulations. The next step includes defining a grid box around the active site of the protein with dimensions of X\u0026thinsp;=\u0026thinsp;157.625, Y\u0026thinsp;=\u0026thinsp;174.369 and Z\u0026thinsp;=\u0026thinsp;218.693 with extension of 30\u0026Aring;. The resulting docking poses and binding affinities were subsequently utilized for molecular dynamics (MD) simulation analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e2.6 MD Simulation analysis\u003c/h2\u003e\u003cp\u003eThe protein-ligand complexes derived from the docking studies were utilized for MD simulation which was carried out with 100ns long run. The detailed methodology for the MD simulation study has been described in some of the previous investigations [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. Brief description is however provided in the Supplementary information to describe the set-up of MD simulation and its analyses.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. RESULTS AND DISCUSSION","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1 2D-QSAR modelling\u003c/h2\u003e\n \u003cp\u003eAs the strategy is outlined before, the 2D-QSAR models were developed with the intention of having better interpretable models without sacrificing the predictive quality of the models. To develop best linear model, two sets of alvaDesc descriptors sets were considered and these are \u003cem\u003e(a)\u003c/em\u003e interpretable descriptors and \u003cem\u003e(b)\u003c/em\u003e all descriptors [\u003cspan class=\"CitationRef\"\u003e62\u003c/span\u003e] to check which set may provide the best solution. Naturally, the chance of getting a better QSAR model is always greater for the latter set since the former is just a subset of the latter. However, due to the fact the feature selection strategy is a complex method, it is always worthwhile to check which set produces the better result. If the second set is indeed found to generate the better results, it is required to check how much difference is found between the best models produced with by these sets. Evidently, a predictive model produced by the interpretable descriptors is capable of providing high statistical quality as well as mechanistic interpretability. Since, it is not wise to compromise the statistical results for sake of molecular interpretability, the current approach clarifies whether inclusion of all descriptors is indeed required.\u003c/p\u003e\n \u003cp\u003eIn this particular investigation, two distinct approaches were conceived to develop the 2D-QSAR models which required utilization of SFS and GA algorithms. The linear regression models were developed from the statistics resulted from the training and test sets. To consider the best predictive model from this investigation the average of \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e; an estimation of the internal predictivity, and \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e; estimating the external predictivity was considered. The findings from the 2D-QSAR analysis are listed in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e [\u003cspan class=\"CitationRef\"\u003e51\u003c/span\u003e].\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSynopsis of the statistical parameters from 2D-QSAR models.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eMethod\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eScore\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eFold\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eInterpretable descriptors\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eAll descriptors\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eQ\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eQ\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.642\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.864\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.828\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.695\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.825\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMAE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.664\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMAE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.689\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.522\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.664\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.833\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.795\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMGD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.678\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.814\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMGD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.683\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.795\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.762\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.208\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.827\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eND\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e#ND: Not Determined\u003c/p\u003e\n \u003cp\u003eSince multiple models were generated here with different descriptor sets and model development strategy, we selected the most predictive model from maximum Q\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e, which justified the internal predictivity of the model. Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e clearly demonstrates that the interpretable descriptors were inadequate to develop a predictive model as predictive as the models produced with all alvaDesc descriptors. The most predictive model was produced when the SFS-MLR model was constructed using \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e as scoring function (without CV). This model (henceforth designated as Model 1) showed \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e of 0.864 and \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e of 0.828. This SFS- MLR model is presented in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e with 2D-QSAR equation and statistical parameters for the training and test sets. Overall, the model is generated with a satisfactory statistical predictivity and it is also evident from statistical parameters other than \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e. Reports suggested that validation parameters based on \u003cem\u003erm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e statistics are more robust than these two parameters [\u003cspan class=\"CitationRef\"\u003e66\u003c/span\u003e]. The internal validation parameters \u003cem\u003erm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTrain\u003c/em\u003e\u003c/sub\u003e and the \u003cem\u003e\u0026Delta;rm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTrain\u003c/em\u003e\u003c/sub\u003e values of the model were found as 0.814 and 0.059, respectively. At the same time, the \u003cem\u003erm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTest\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003e∆rm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTest\u003c/em\u003e\u003c/sub\u003e values are 0.769 and 0.120, respectively. The observed vs. predicted pIC\u003csub\u003e50\u003c/sub\u003e values of 2D-QSAR model is depicted in Fig. 1.\u003c/p\u003e\n \u003cp\u003eAfter that we checked the correlation matrix of this model (shown in Fig. 1) and it was observed that these descriptors are devoid of high inter-collinearity as the highest absolute correlation (Pearson \u003cem\u003er\u003c/em\u003e) between any two descriptor is 0.45 (i.e., \u0026lt; 0.95).\u003c/p\u003e\n \u003cp\u003eY-randomization was performed by shuffling the pIC50 values of the compounds over 1000 iterations, yielding a \u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eP\u003c/em\u003e\u003c/sub\u003e value of 0.837. The relatively high score confirms the statistical reliability of the model and not generated by random chance.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eDetailed statistical parameters of 2D-QSAR model.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eEquation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining set\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTest set\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"7\"\u003e\n \u003cp\u003e\u003cem\u003epIC\u003c/em\u003e\u003csub\u003e\u003cem\u003e50\u003c/em\u003e\u003c/sub\u003e=-6.404(\u0026plusmn;\u0026thinsp;0.844) *\u003cem\u003eVE1sign_B(m)\u003c/em\u003e -2.456(\u0026plusmn;\u0026thinsp;0.595) *\u003cem\u003eGATS8v\u003c/em\u003e\u0026thinsp;+\u0026thinsp;3.866(\u0026plusmn;\u0026thinsp;0.729) *\u003cem\u003eTDB03p\u003c/em\u003e\u0026thinsp;+\u0026thinsp;1.815(\u0026plusmn;\u0026thinsp;0.28) *\u003cem\u003eL3v\u003c/em\u003e \u0026minus;\u0026thinsp;0.333(\u0026plusmn;\u0026thinsp;0.066) *\u003cem\u003eF02[C-N]\u003c/em\u003e\u0026thinsp;+\u0026thinsp;0.257(\u0026plusmn;\u0026thinsp;0.051) *\u003cem\u003eCATS3D_02_DL\u003c/em\u003e\u0026thinsp;+\u0026thinsp;13.555(\u0026plusmn;\u0026thinsp;3.049) *\u003cem\u003eWHALES20_IR\u003c/em\u003e\u0026thinsp;+\u0026thinsp;5.46(\u0026plusmn;\u0026thinsp;1.756)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003etr\u003c/em\u003e\u003c/sub\u003e =59, \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.900,\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003ets\u003c/em\u003e\u003c/sub\u003e = 14, \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e/Q\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eF1\u003c/em\u003e\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eadj\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.886,\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e=\u0026thinsp;0.828, \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eF2\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.830,\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.864,\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eF3\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.848,\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eMAE\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.029,\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eRMSEP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.420,\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMSE\u0026thinsp;=\u0026thinsp;0.341\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003erm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTest\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.769,\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003erm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTrain\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.814,\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e∆rm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTest\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.120\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e∆rm\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eTrain\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eMoreover, it is absolutely important to check the applicability domain (AD) for the establishment of the 2D-QSAR model. The Williams plot was generated for the Model 1, which is presented in the Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. No structural or response outliers were found in this model. However, standardization approach detected one structural outlier in the training set and we retained this since no test set compound was identified as outlier.\u003c/p\u003e\n \u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e(B) provides an idea about the impact of each descriptor from the model M1. The model with the all descriptors shows insightful structural information of 6‑Anilino Imidazo[4,5‑c] pyridin-2-one derivatives. At the same time, out of 7 descriptors, two (\u003cem\u003ei.e., WHALES20_IR\u003c/em\u003e and \u003cem\u003eVE1sign_B(m)\u003c/em\u003e) showed high significance and three (\u003cem\u003ei.e., TDB03p\u003c/em\u003e, \u003cem\u003eL3v\u003c/em\u003e and \u003cem\u003eGATS8v\u003c/em\u003e) demonstrated moderate significance. Two descriptors, namely \u003cem\u003eCATS3D_02_DL and F02[C-N]\u003c/em\u003e depicted negligible significance. The descriptions of the descriptors of the model are outlined in Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. Noticeably, most of the descriptors of the model is topological in nature and less interpretable. We first describe the \u003cem\u003eWHALES20_IR\u003c/em\u003e descriptor which is found to be the most influential descriptor from Model 1 and it was found to have positive correlation towards the response variable. Weighted Holistic Atom Localization and Entity Shape (\u003cem\u003eWHALES\u003c/em\u003e) descriptors are kind of 3D topological descriptors, which capture molecular shape and partial charges concurrently. Here IR of the \u003cem\u003eWHALES20_IR\u003c/em\u003e denotes isolation and remoteness, and 20 is ratio percentile. Isolation depends on the local electron density and nearby significant group; remoteness is influenced by the topological distance in 3D structure. Molecules with different IR profiles may have different impact on the biological activity whereas 20 percentile value focuses on the less isolated regions, which is critical for the active site for the ligands. Overall, it can be concluded that the specific 3D shape of the compounds is important to determine the biological activity. The maximum value of this descriptor is found in \u003cstrong\u003e88\u003c/strong\u003e, which is also the most potent compound of the dataset with reported pIC\u003csub\u003e50\u003c/sub\u003e value of 8.674. On the other hand, another compound \u003cstrong\u003e29\u003c/strong\u003e that showed the lowest value of \u003cem\u003eWHALES20_IR\u003c/em\u003e (i.e., -0.135) was found to possess an pIC\u003csub\u003e50\u003c/sub\u003e of 5.442. In Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, we demonstrated and compared the structures of the most potent compound \u003cstrong\u003e88\u003c/strong\u003e and least potent compound \u003cstrong\u003e41\u003c/strong\u003e, which has \u003cem\u003eWHALES20_IR\u003c/em\u003e value of -0.108. In the same figure, similar comparisons were performed with second and third most influential descriptors - \u003cem\u003eVE1sign_B(m)\u003c/em\u003e and \u003cem\u003eTDB03p\u003c/em\u003e. The second most important descriptor is \u003cem\u003eVE1sign_B(m)\u003c/em\u003e, with the negative corelation with the response variable. It is a 2D matrix-based descriptors that is weighted with atomic mass of the compounds. The third most important descriptor is \u003cem\u003eTDB03p\u003c/em\u003e, under 3D autocorrelations category. This depends on the polarizability of the compounds, as it is positively corelated with the response variable the higher polarizability enhances the binding potential of the compounds which is shown in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. Interestingly, the most potent compound \u003cstrong\u003e88\u003c/strong\u003e was found to have maximum value (i.e., 1.831) of this descriptor whereas the lowest descriptor value (i.e., 1.435) was found in a low active derivative \u003cstrong\u003e73\u003c/strong\u003e (pIC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;5.714). Two least potent derivatives \u003cstrong\u003e20\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e had \u003cem\u003eTDB03p\u003c/em\u003e values of 1.486 and 1.479, respectively. In comparison to \u003cstrong\u003e88\u003c/strong\u003e, all these low active compounds contained more bulky non-polar residues such as cyclopentyl, that may be responsible for low activity towards PKI enzyme.\u003c/p\u003e\n \u003cp\u003eThe fourth and the fifth most important descriptors are \u003cem\u003eGATS8v\u003c/em\u003e and \u003cem\u003eL3v\u003c/em\u003e respectively. Even though \u003cem\u003eGATS8v\u003c/em\u003e belongs to the 2D autocorrelations category and \u003cem\u003eL3v\u003c/em\u003e is a \u003cem\u003eWHIM\u003c/em\u003e descriptor, both these are weighted by the van der Waals volume and both of them are found to be positively correlated with the response variable. Both descriptors bank on the van der Waals volume. Two descriptors with least significance are \u003cem\u003eF02[C-N]\u003c/em\u003e and \u003cem\u003eCATS3D_02_DL\u003c/em\u003e. The \u003cem\u003eF02[C-N]\u003c/em\u003e belongs to 2D-atom pair descriptor which denotes the depicts the number of times the pair C and N appear within the molecular structure at topological distance of 2. Low value of this descriptor was found to be important. Finally, CATS descriptors are pharmacophore descriptors and \u003cem\u003eCATS3D_02_DL\u003c/em\u003e quantifies the frequency of donor-lipophilic feature pairs at a defined 3D distance interval (bin 02). Positive correlation of this particular descriptor with the response variable signifies that the presence of this structural feature is beneficial to the binding potential of the compounds [\u003cspan class=\"CitationRef\"\u003e67\u003c/span\u003e]. Detailed description of each molecular descriptor in the developed model has been provided in the Table S2 of Supplementary information.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2 MolSHAP analysis\u003c/h2\u003e\n \u003cp\u003e2D-QSAR model provided satisfactory statistical results but provided us limited information about the contributions of various substitutions for increased or decreased potential towards DNA-PK. As it has been discussed earlier, MolSHAP is a comparatively novel strategy of QSAR model development that allows \u003cem\u003eR\u003c/em\u003e-group analyses and may be considered as upgraded strategy of Free Wilson analysis. Here, the model generated with Gaussian Process Regressor (GPR) showed \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e values of 0.987 and 0.844, respectively. However, it was necessary to more robust internal validation parameter and for this 5-fold cross-validation was performed with the training set that resulted in \u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e value of 0.583. Therefore, the model depicted satisfactory statistical predictivity. First, the contributions of the \u003cem\u003eR\u003c/em\u003e groups were analysed and the results are depicted in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. In this figure the core common structure is shown and 10 most favourable and 10 most unfavourable fragments are provided.\u003c/p\u003e\n \u003cp\u003eThe \u003cem\u003eR\u003c/em\u003e-group analyses revealed that the \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e7\u003c/em\u003e\u003c/sub\u003e positions (both attached to imidazo[4,5-c]- pyridine-2-one moiety) are the most important regions for determining increased biological activity. The presence of 4-methoxyphenyl moiety at position \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sub\u003e of the common structure appears as the most contributing fragment for increasing biological activity and it is followed by 4-hydroxyphenyl and 4-(2-methyl-2-phenylpropanenitrile) moieties. Noticeably, the latter fragments are found in two most potent compounds; \u003cstrong\u003e87\u003c/strong\u003e and \u003cstrong\u003e88\u003c/strong\u003e, with pIC\u003csub\u003e50\u003c/sub\u003e values of 8.357 and 8.674 respectively. Substitutions at positions \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e4\u003c/em\u003e\u003c/sub\u003e also determine improved activity. Among this, the presence of methyl group at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e exhibits maximum impact with contribution of 0.83. Noticeably, the presence of isopropyl group at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e7\u003c/em\u003e\u003c/sub\u003e is responsible for lowering the biological potency. Noticeably, the presence of this moiety is found in \u003cstrong\u003e72\u003c/strong\u003e that has pIC\u003csub\u003e50\u003c/sub\u003e value 5.975. However, replacement of this isopropyl moiety with methyl group gives rise to \u003cstrong\u003e53\u003c/strong\u003e with significant improvement of biological activity (pIC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;8.134). The MolSHAP analysis clearly indicates that the presence of methyl group at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e4\u003c/em\u003e\u003c/sub\u003e is favourable whereas methyl group at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sub\u003e is detrimental for increased potency. Similarly, influence of cyclopentyl was found to vary depending on its position. Substitution at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sub\u003e was favourable whereas \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e7\u003c/em\u003e\u003c/sub\u003e in unfavourable. Overall, the 2D-QSAR analysis highlights the favourable and unfavourable fragments but does not indicate the electrostatic and steric factors responsible for the increased or decreased potentials towards PKI enzyme. Therefore, we resorted to 3D-QSAR analyses to extract more information regarding these.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 3D- QSAR modelling\u003c/h2\u003e\n \u003cp\u003eAs seen in Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, the better predictive quality was achieved through the deployment of FFD-SEL algorithm. The number of ligands considered to for training set was 59 while 14 datapoints were set aside for test set prediction. The \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLOO\u003c/em\u003e\u003c/sub\u003e (0.726), \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLTO\u003c/em\u003e\u003c/sub\u003e (0.760) and \u003cem\u003eQ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003eLMO\u003c/em\u003e\u003c/sub\u003e (0.736) values indicates that the model has strong internal predictivity, as well as external predictability (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003csub\u003e\u003cem\u003ePred\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.827). Additionally, the 3D-QSAR model accurately captures the role of both steric as well as the electrostatic fields in model development. Figure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e shows a visual depiction of the contour maps.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eStatistical parameters from 3D-QSAR analysis\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFFD-SEL\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUVE-PLS\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN\u003csub\u003etraining\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e59\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eF-test\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e134.456\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e125.997\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e/SDEC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.908/0.321\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.903/ 0.331\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eQ\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLOO\u003c/sub\u003e/SDEP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.762/ 0.518\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.695/0.588\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eQ\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLTO\u003c/sub\u003e/SDEP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.760/0.521\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.691/0.591\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eQ\u003csup\u003e2\u003c/sup\u003e\u003csub\u003eLMO\u003c/sub\u003e/SDEP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.736/0.546\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.658/ 0.622\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN\u003csub\u003etest\u003c/sub\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003csub\u003ePred\u003c/sub\u003e/SDEP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.827/0.446\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.786/0.495\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe contributions of the steric and electrostatic fields were found as 58% and 42%, respectively indicating the joint responsibility of both components in shaping the model. There is substantial consistency between the MolSHAP analyses and 3D-QSAR modelling.\u003c/p\u003e\n \u003cp\u003eThe presence of relatively bulky and electron rich 4- 2-methyl-2-phenylpropanenitrile moiety at \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sub\u003e-position was inserted into a steric favourable as well as electronegative fields. At the same position (i.e., \u003cem\u003eR\u003c/em\u003e\u003csub\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sub\u003e), the lowest active \u003cstrong\u003e41\u003c/strong\u003e contained relatively less electron rich cyclopentyl ring that was inserted into an electronegative field. Being a bulky group, the cyclopentyl was found close to neither steric favourable nor steric unfavourable fields.\u003c/p\u003e\n \u003cp\u003eCompound \u003cstrong\u003e88\u003c/strong\u003e, the one with the highest binding affinity contains a bulky aromatic side chain that aligns well within the steric favourable region of the isocontour map. In addition, a methoxy and a methyl substituent are also positioned in sterically advantageous regions, prompting further enhancement in its binding potential. By contrast, compound \u003cstrong\u003e41\u003c/strong\u003e; the one with the weakest binding affinity, lacks any bulky substituents near the steric favourable region, explaining one of the causes of its weak binding potential. Analysis of the electrostatic isocontour maps also revealed that compound \u003cstrong\u003e88\u003c/strong\u003e places its electron rich aromatic ring in closer proximity to an electrostatically favourable region, supporting beneficial interactions. On the other hand, compound \u003cstrong\u003e41\u003c/strong\u003e contains a hydrophobic cyclopentyl moiety within the same electronegative favourable region. Since non-polar groups are unable to interact effectively in such regions, this unfavourable placement likely results in decreased binding affinity. Overall, these observations provide important insights into the structure-activity relationship of DNA-dependent protein kinase inhibitors.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4 Molecular dynamics simulation\u003c/h2\u003e\n \u003cp\u003eThe 2D-QSAR, MolSHAP and 3D-QSAR provided various information about the structural attributes for dataset compounds for enhanced potential regarding PKI. Nevertheless, it was necessary to understand the dynamic behaviours of the 6‑Anilino Imidazo[4,5‑c] pyridin-2-one\u0026rsquo;s derivatives with PKI. In order to understand this, we selected the most active compound \u003cstrong\u003e88\u003c/strong\u003e and one of the least active compounds \u003cstrong\u003e41\u003c/strong\u003e for MD simulation analyses. For molecular docking the protein structure provided by Hong et al was used in the current work. Notably, they also provided the docked complexes of Dactolisib and another dataset compound \u003cstrong\u003e78\u003c/strong\u003e. In the current work, we used the AZD7648-PKI complex as a reference to compare the dynamic behaviours of \u003cstrong\u003e88\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e. Interestingly, when the molecular docking was performed with \u003cstrong\u003e88\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e using Autodock Vina, the best docked pose of \u003cstrong\u003e88\u003c/strong\u003e replicated the structure of both AZD7648 as well as \u003cstrong\u003e78\u003c/strong\u003e (both provided by Hong et al). However, the docked conformation of \u003cstrong\u003e41\u003c/strong\u003e failed to replicate the similar conformation that led us to perform forced alignment of \u003cstrong\u003e41\u003c/strong\u003e with the docked pose of \u003cstrong\u003e78\u003c/strong\u003e to check the reason for this (Fig. \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e). For this alignment, the .\u003cem\u003esdf\u003c/em\u003e form of \u003cstrong\u003e78\u003c/strong\u003e was used as a template whereas the structure of \u003cstrong\u003e41\u003c/strong\u003e was provided as SMILES. Subsequently, the aligned structure of \u003cstrong\u003e41\u003c/strong\u003e was minimized by \u003cem\u003eOpenBabel\u003c/em\u003e tool using MMFF94 forcefield. The minimized structure of \u003cstrong\u003e41\u003c/strong\u003e was found to have multiple clashes at the binding site (shown in Fig. \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e) that led us to proceed with the docked conformation of both \u003cstrong\u003e88\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e for 100ns MD simulations. After the MD simulation, we performed trajectory analyses, MM-GBSA analyses and per-residue decomposition analyses. The MM-GBSA results that are provided in Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e match the differences in experimental biological activities of \u003cstrong\u003e88\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eMM-GBSA binding energies (in kcal/mol) from MD simulations\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eComplexes\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Delta;E\u003csub\u003evdW\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Delta;E\u003csub\u003eelec\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Delta;G\u003csub\u003epolar\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Delta;G\u003csub\u003enonpolar\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eT\u0026Delta;S\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Delta;G\u003csub\u003ebind(T)\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003epIC\u003csub\u003e50\u003c/sub\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e88\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-54.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-7.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-6.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-22.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-25.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.674\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e41\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-44.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-6.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-5.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-29.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-10.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAZD7648\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-47.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-7.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-4.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-18.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-23.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eMore importantly, the best active 88 depicted significantly increased van der Waals interactions and considerably higher hydrogen bond interactions than low active \u003cstrong\u003e41\u003c/strong\u003e. Even though the nonpolar solvation energy of \u003cstrong\u003e88\u003c/strong\u003e was found to be more favourable, the polar solvation energy of the compound was less favourable for 88 that may be due to its larger molecular weight. The trajectory analyses, performed with \u003cstrong\u003e88\u003c/strong\u003e, \u003cstrong\u003e41\u003c/strong\u003e and \u003cstrong\u003eAZD7684\u003c/strong\u003e revealed sufficient dynamic stabilities of all these ligands during MD simulations. However, slightly increased fluctuations were noted for low active \u003cstrong\u003e41\u003c/strong\u003e (average RMSF\u0026thinsp;=\u0026thinsp;1.66 \u0026Aring;) as compared to \u003cstrong\u003e88\u003c/strong\u003e (average RMSF\u0026thinsp;=\u0026thinsp;1.55 \u0026Aring;). Furthermore, as shown in Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e, 88 demonstrated higher number of hydrogen bond interactions in comparison with \u003cstrong\u003e41\u003c/strong\u003e further confirming that the former may produce increased electrostatic interactions with the receptor.\u003c/p\u003e\n \u003cp\u003eIn Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e, the last poses of these two compounds are depicted. However, to understand more specifically about the interactions between these ligands and DNA-PK, we performed the per residue decomposition analyses with the amino acids located within 10\u0026Aring; radius of the ligand.\u003c/p\u003e\n \u003cp\u003eNoticeably, when we compare the per residue decomposition analyses results of \u003cstrong\u003e88\u003c/strong\u003e and \u003cstrong\u003e41\u003c/strong\u003e, the former depicted increased interactions with the receptor. In one hand, the electrostatic interactions of \u003cstrong\u003e88\u003c/strong\u003e with Lys3753 and Leu3806 were found to be significantly greater than \u003cstrong\u003e41\u003c/strong\u003e whereas van der Waals interaction of \u003cstrong\u003e88\u003c/strong\u003e with Trp3806 was found to be predominant in determining its higher binding potential. In this regard, we analysed the hydrogen bonds formed between these two ligands and the receptor. The trajectory analyses also revealed that the hydrogen bond interactions between pyridine nitrogen of \u003cstrong\u003e88\u003c/strong\u003e and Leu3806 may play the most significant role. Apart from that, the oxygen atom of the methoxy group of \u003cstrong\u003e88\u003c/strong\u003e forms hydrogen bond interactions with Lys3753. Compound \u003cstrong\u003e41\u003c/strong\u003e fails to show such significant interactions with the receptor and in this case, it is the Ser3731 that forms maximum hydrogen bond interaction with this compound. Noticeably, some \u0026pi;- \u0026pi; interactions were found to play pivotal role in determining the binding affinity. Interestingly, trajectory analyses revealed that almost 97% frames were identified in the last 50 ns MD simulation run where \u0026pi;- \u0026pi; interactions were found between heteroaromatic residues of \u003cstrong\u003e88\u003c/strong\u003e and Trp3805. However, only negligible number of frames were found where the aromatic rings of \u003cstrong\u003e41\u003c/strong\u003e engaged in \u0026pi;- \u0026pi; interactions with Trp3805 and Tyr3791. All these clearly indicate that it is the overall 3D topology of \u003cstrong\u003e41\u003c/strong\u003e that may hinder it to mimic the binding conformation adopted by \u003cstrong\u003e88\u003c/strong\u003e and other higher active compounds and it ultimately leads to lower activity towards DNA-PK.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\n \u003ch2\u003e3.5 MolSHAP based novel compound design\u003c/h2\u003e\n \u003cp\u003ePreviously, we discussed development predictive MolSHAP QSAR modelling to understand favourable and unfavourable fragments. However, the method is also used to design and propose new compounds. Here, total 239 compounds were optimized and ranked as per the predicted pIC\u003csub\u003e50\u003c/sub\u003e values. We picked six structures with predicted pIC\u003csub\u003e50\u003c/sub\u003e greater than \u003cstrong\u003e88\u003c/strong\u003e (the most active compound of the dataset) and their structures were found that are depicted in Fig. \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eHowever, these MolSHAP predicted structures were screened with the 2D-QSAR model, which was so far proved to be the most predictive model, MS02 and MS06 were found as structural outliers by the standardization method. Significantly, the molecular docking performed with these six compounds also depicted that except these structural outliers, the best poses of remaining four MolSHAP predicted compounds superimposed well with the bound conformation of AZD or docked pose of \u003cstrong\u003e78\u003c/strong\u003e (Provided by Hong et al). Now, after screening with our 2D-QSAR model, these four compounds exhibited predicted pIC\u003csub\u003e50\u003c/sub\u003e values as follows: MS01: 8.593, MS03: 8.397, MS04: 7.686 and MS05: 8.079. Therefore, as per the predictions provided by both MolSHAP and 2D-QSAR models, MS01 emerged as the most promising structure designed by MolSHAP. In order to ensure further about its reliability, we performed 100ns MD simulations with the docked conformation of it. This compound depicted \u0026Delta;G\u003csub\u003ebind(T)\u003c/sub\u003e of -20.43 kCal/mol, which is very close to AZD7684. The energy contributions are as follows: \u0026Delta;E\u003csub\u003evdW\u003c/sub\u003e = -48.03 kCal/mol, \u0026Delta;E\u003csub\u003eelec\u003c/sub\u003e = -8.47 kCal/mol, \u0026Delta;G\u003csub\u003epolar\u003c/sub\u003e = 19.48 kCal/mol, \u0026Delta;G\u003csub\u003enonpolar\u003c/sub\u003e = -5.46 kCal/mol and T\u0026Delta;S = -22.04 kCal/mol. Noticeably, this compound has highly satisfactory polar interactions that is even greater than that observed for \u003cstrong\u003e88\u003c/strong\u003e. Additionally, trajectory analyses (shown in Supplementary information) depicted that MS01 maintains conformational stability and forms 5\u0026ndash;7 hydrogen bond interactions during MD run.\u003c/p\u003e\n \u003cp\u003eAnalyses of the last pose of MS01 after MD simulation and the results of per residue decomposition analyses hint that MS01 may form stable hydrogen bond interactions with Leu3806 and Glu3804. Furthermore, hydrogen bond analyses indicated that its methoxy residue may also form stable hydrogen bond interactions with Asn3926. Similar to \u003cstrong\u003e88\u003c/strong\u003e, it also shows stable \u0026pi;-\u0026pi; interactions with Trp3805 as around 99% of frames in the last 50ns MD run depicted its existence and according to per residue decomposition analyses, van der Waals interactions with this amino acid residue may play predominant role in its binding with the receptor as shown in Fig. \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. CONCLUSION","content":"\u003cp\u003eDNA-PK inhibitors have emerged as one of the leading candidates as radiosensitizers and the Hong et al designed a series of 6‑Anilino Imidazo[4,5‑c]pyridin-2-one derivatives through the much-coveted scaffold hopping strategy. This particular investigation, seeking to unravel the structural requirements of the DNA-PK inhibitors, constitutes of performing 2D-QSAR and 3D-QSAR analysis. Apart from the conventional QSAR analyses, the study also focuses on understanding the contributions of different structural fragments of the dataset ligands. This is followed by executing molecular dynamics simulation analysis for ligands with greatest and lowest biological activities. Finally, with the help of MolSHAP, new ligands were designed and the ligand with the best predicted activity was subjected to MD simulation analysis to reveal the various molecular interactions while binding with the receptor protein.\u003c/p\u003e\n\u003cp\u003eFirstly, the 2D-QSAR analysis retrieved satisfactory results with both internal and external predictivities of 0.886 and 0.828 respectively. Furthermore, the model also brought forward the most significant descriptors from the dataset. It was found out that topological descriptor (WHALES20_IR) leverages a crucial influence on the binding potential of the ligands. The 3D topology of the structures provides an important idea about the electron density of the various atoms in the structure which is relevant in exerting binding capabilities of the ligands. Subsequently, the presence of bulky steric groups may influence negatively to the pIC\u003csub\u003e50\u003c/sub\u003e of the ligands. Descriptors from the 2D-QSAR analyses strongly indicate the importance of partial charges and van der Waals volume, exerting key influence on the binding potential. This was further supported by the MolSHAP analysis which clearly states the contributions of different fragments of the compounds in the dataset. From the MolSHAP analysis, it was found out that 4-methoxyphenyl moiety had the highest positive contribution in showing greater binding potential. The greater electron density of the group holds a significant importance that is also backed up by the findings from the 2D-QSAR analyses. Similar effects can be observed for the hydroxyphenyl moiety which is a strong electron donating group, having higher density of electrons in the structure. This phenomenon was also observed with 2D-QSAR analyses which heavily relied on the contributions of 3D topology and presence of electron rich regions of the dataset ligands.\u003c/p\u003e\n\u003cp\u003eThe findings from the 3D-QSAR analyses were found to be consistent with the information extracted from MolSHAP analysis which indicated the significance of electron rich 4-methoxyphenyl group in shaping the binding potential of the ligands. The 3D-QSAR isocontour maps also provides evidence of the electronegative fragment of the compound \u003cstrong\u003e88\u003c/strong\u003e being inserted into the electrostatic favourable region. The significance of the fragment was further consolidated by the fact that due to its bulky nature, it was also present in the steric favourable region of the isocontour map. On the other hand, for compounds with lesser binding affinity, the less electronegative fragment like cyclopentyl moiety was in close proximity to the electronegative region of the isocontour map. The consistency between the findings from the 2D, 3D and MolSHAP analyses encouraged us into looking at findings from the MD simulations. To understand the interactions between the ligands and receptor, 100ns MD simulation analyses were employed which gave a clear indication of the participating interactive force between them. The MM-GBSA analysis for both the highest and lowest active compounds along with the bound ligand\u0026nbsp;AZD7684 provided insights on how electrostatic interactions might play an important role in determining the binding potential of the ligands. The trajectory analysis revealed that compound \u003cstrong\u003e88\u003c/strong\u003e participated in more hydrogen bond interactions than \u003cstrong\u003e41\u003c/strong\u003e which leads to its greater binding potential. The per-residue decomposition analysis provided evidence of electrostatic interactions between \u003cstrong\u003e88\u003c/strong\u003e and amino acid residues Lys3753 and Leu3806; while the van der Waals interactions with Trp3806 was the predominant factor contributing to the binding affinity of the ligand. These interactions were either absent or very weak in case of compound \u003cstrong\u003e41\u003c/strong\u003e; leading to its lower biological potential. Finally, it was revealed that π-π interaction between the heterocyclic structure of 88 and Trp3805 was captured in almost 97% frames of the last 50ns of the MD run which was scarce for compound \u003cstrong\u003e41\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eUpon understanding the ligand-receptor interaction, the novel designed molecule through MolSHAP was subjected to MD simulation analysis. The best predictive compound \u003cstrong\u003eMS01\u003c/strong\u003e was found to exert similar MM-GBSA analysis to that of AZD7684 and the polar interactions were found to be more potent than that of \u003cstrong\u003e88\u003c/strong\u003e. The per-residue analysis of the ligand and receptor revealed stable H-bond interactions similar to \u003cstrong\u003e88\u003c/strong\u003e. Furthermore, \u003cstrong\u003eMS01\u003c/strong\u003e also exhibited π-π interactions with Trp3805 which was found to be a crucial factor in exerting binding potential of the compounds. The findings from this investigation can be of enormous help in designing new DNA dependent protein kinase inhibitors as the structural requirements along with the relevant interactions with the receptor amino acids have been explored in details. Furthermore, the entire work is based on open-access tools that require no subscription cost and for this, the models are reproducible.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCRediT authorship contribution statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eS. Mitra:\u003c/strong\u003e Writing – original draft, Investigation, Formal analysis, Validation, Resources, Data curation, Software. \u003cstrong\u003eR. K. Dolai:\u003c/strong\u003e Investigation, Data curation, Writing – original draft, Software. \u003cstrong\u003eN. Ghosh:\u003c/strong\u003e Writing – review \u0026amp; editing, Formal analysis, Conceptualisation, Methodology, Funding acquisition, Investigation, Supervision. \u003cstrong\u003eS. C. Mandal:\u003c/strong\u003e Formal analysis, Validation, Resources. \u003cstrong\u003eA. K. Halder:\u003c/strong\u003e Writing – review \u0026amp; editing, Conceptualisation, Visualisation, Validation, Supervision, Investigation, Formal analysis, Software.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis current work was funded by the Department of Science and Technology and Biotechnology, Govt. of West Bengal, India Vide Memo. 2027 (Sanc.)/STBT-11012 (19)/ 6/2023-ST SEC, dated 24–01-2024.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is provided within the manuscript or supplementary information files\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eDelaney G, Jacob S, Featherstone C, Barton M (2005) The role of radiotherapy in cancer treatment: estimating optimal utilization from a review of evidence-based clinical guidelines. Cancer 104:1129-37. https://doi.org/10.1002/cncr.21324.\u003c/li\u003e\n\u003cli\u003eModing EJ, Kastan MB, Kirsch DG (2013) Strategies for optimizing the response of cancer and normal tissues to radiation. Nat Rev Drug Discov 12:526-42. https://doi.org/10.1038/nrd4003.\u003c/li\u003e\n\u003cli\u003ePignon JP, Bourhis J, Domenge C, Designe L (2000) Chemotherapy added to locoregional treatment for head and neck squamous-cell carcinoma: three meta-analyses of updated individual data. MACH-NC Collaborative Group. Meta-Analysis of Chemotherapy on Head and Neck Cancer. Lancet 355:949-55.\u003c/li\u003e\n\u003cli\u003eBentzen SM, Trotti A (2007) Evaluation of early and late toxicities in chemoradiation trials. J Clin Oncol 25:4096-103. https://doi.org/10.1200/JCO.2007.13.3983.\u003c/li\u003e\n\u003cli\u003eEwing D (1998) The Oxygen Fixation Hypothesis: A Reevaluation. 21:355-61.\u003c/li\u003e\n\u003cli\u003eJing X, Yang F, Shao C, Wei K, Xie M, Shen H, et al. (2019) Role of hypoxia in cancer therapy by regulating the tumor microenvironment. Mol Cancer 18:157. https://doi.org/10.1186/s12943-019-1089-9.\u003c/li\u003e\n\u003cli\u003eWilson WR, Hay MP (2011) Targeting hypoxia in cancer therapy. Nat Rev Cancer 11:393-410. https://doi.org/10.1038/nrc3064.\u003c/li\u003e\n\u003cli\u003eOvergaard J (2007) Hypoxic radiosensitization: adored and ignored. J Clin Oncol 25:4066-74. https://doi.org/10.1200/JCO.2007.12.7878.\u003c/li\u003e\n\u003cli\u003eLiew LP, Shome A, Wong WW, Hong CR, Hicks KO, Jamieson SMF, et al. (2023) Design, Synthesis and Anticancer Evaluation of Nitroimidazole Radiosensitisers. Molecules 28. https://doi.org/10.3390/molecules28114457.\u003c/li\u003e\n\u003cli\u003eGroelly FJ, Fawkes M, Dagg RA, Blackford AN, Tarsounas M (2023) Targeting DNA damage response pathways in cancer. Nat Rev Cancer 23:78-94. https://doi.org/10.1038/s41568-022-00535-5.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Connor MJ (2015) Targeting the DNA Damage Response in Cancer. Mol Cell 60:547-60. https://doi.org/10.1016/j.molcel.2015.10.040.\u003c/li\u003e\n\u003cli\u003ePaques F, Haber JE (1999) Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae. Microbiol Mol Biol Rev 63:349-404. https://doi.org/10.1128/MMBR.63.2.349-404.1999.\u003c/li\u003e\n\u003cli\u003eMahaney BL, Meek K, Lees-Miller SP (2009) Repair of ionizing radiation-induced DNA double-strand breaks by non-homologous end-joining. Biochem J 417:639-50. https://doi.org/10.1042/BJ20080413.\u003c/li\u003e\n\u003cli\u003eScully R, Panday A, Elango R, Willis NA (2019) DNA double-strand break repair-pathway choice in somatic mammalian cells. Nat Rev Mol Cell Biol 20:698-714. https://doi.org/10.1038/s41580-019-0152-0.\u003c/li\u003e\n\u003cli\u003eShao Z, Davis AJ, Fattah KR, So S, Sun J, Lee KJ, et al. (2012) Persistently bound Ku at DNA ends attenuates DNA end resection and homologous recombination. DNA Repair (Amst) 11:310-6. https://doi.org/10.1016/j.dnarep.2011.12.007.\u003c/li\u003e\n\u003cli\u003eDowns JA, Jackson SP (2004) A means to a DNA end: the many roles of Ku. Nat Rev Mol Cell Biol 5:367-78. https://doi.org/10.1038/nrm1367.\u003c/li\u003e\n\u003cli\u003eMimori T, Hardin JA (1986) Mechanism of interaction between Ku protein and DNA. J Biol Chem 261:10375-9.\u003c/li\u003e\n\u003cli\u003eHartley KO, Gell D, Smith GC, Zhang H, Divecha N, Connelly MA, et al. (1995) DNA-dependent protein kinase catalytic subunit: a relative of phosphatidylinositol 3-kinase and the ataxia telangiectasia gene product. Cell 82:849-56. https://doi.org/10.1016/0092-8674(95)90482-4.\u003c/li\u003e\n\u003cli\u003eDing Q, Reddy YV, Wang W, Woods T, Douglas P, Ramsden DA, et al. (2003) Autophosphorylation of the catalytic subunit of the DNA-dependent protein kinase is required for efficient end processing during DNA double-strand break repair. Mol Cell Biol 23:5836-48. https://doi.org/10.1128/MCB.23.16.5836-5848.2003.\u003c/li\u003e\n\u003cli\u003eMahajan KN, Nick McElhinny SA, Mitchell BS, Ramsden DA (2002) Association of DNA polymerase mu (pol mu) with Ku and ligase IV: role for pol mu in end-joining double-strand break repair. Mol Cell Biol 22:5194-202. https://doi.org/10.1128/MCB.22.14.5194-5202.2002.\u003c/li\u003e\n\u003cli\u003eNiewolik D, Pannicke U, Lu H, Ma Y, Wang LC, Kulesza P, et al. (2006) DNA-PKcs dependence of Artemis endonucleolytic activity, differences between hairpins and 5\u0026apos; or 3\u0026apos; overhangs. J Biol Chem 281:33900-9. https://doi.org/10.1074/jbc.M606023200.\u003c/li\u003e\n\u003cli\u003eNiewolik D, Peter I, Butscher C, Schwarz K (2017) Autoinhibition of the Nuclease ARTEMIS Is Mediated by a Physical Interaction between Its Catalytic and C-terminal Domains. J Biol Chem 292:3351-65. https://doi.org/10.1074/jbc.M116.770461.\u003c/li\u003e\n\u003cli\u003eFulop GM, Phillips RA (1990) The scid mutation in mice causes a general defect in DNA repair. Nature 347:479-82. https://doi.org/10.1038/347479a0.\u003c/li\u003e\n\u003cli\u003eBiedermann KA, Sun JR, Giaccia AJ, Tosto LM, Brown JM (1991) scid mutation in mice confers hypersensitivity to ionizing radiation and a deficiency in DNA double-strand break repair. Proc Natl Acad Sci U S A 88:1394-7. https://doi.org/10.1073/pnas.88.4.1394.\u003c/li\u003e\n\u003cli\u003eCurtin NJ (2012) DNA repair dysregulation from cancer driver to therapeutic target. Nat Rev Cancer 12:801-17. https://doi.org/10.1038/nrc3399.\u003c/li\u003e\n\u003cli\u003eDavidson D, Amrein L, Panasci L, Aloyz R (2013) Small Molecules, Inhibitors of DNA-PK, Targeting DNA Repair, and Beyond. Front Pharmacol 4:5. https://doi.org/10.3389/fphar.2013.00005.\u003c/li\u003e\n\u003cli\u003eLee TW, Wong WW, Dickson BD, Lipert B, Cheng GJ, Hunter FW, et al. (2019) Radiosensitization of head and neck squamous cell carcinoma lines by DNA-PK inhibitors is more effective than PARP-1 inhibition and is enhanced by SLFN11 and hypoxia. Int J Radiat Biol 95:1597-612. https://doi.org/10.1080/09553002.2019.1664787.\u003c/li\u003e\n\u003cli\u003eSibanda BL, Chirgadze DY, Blundell TL (2010) Crystal structure of DNA-PKcs reveals a large open-ring cradle comprised of HEAT repeats. Nature 463:118-21. https://doi.org/10.1038/nature08648.\u003c/li\u003e\n\u003cli\u003eSibanda BL, Chirgadze DY, Ascher DB, Blundell TL (2017) DNA-PKcs structure suggests an allosteric mechanism modulating DNA double-strand break repair. Science 355:520-4. https://doi.org/10.1126/science.aak9654.\u003c/li\u003e\n\u003cli\u003eYin X, Liu M, Tian Y, Wang J, Xu Y (2017) Cryo-EM structure of human DNA-PK holoenzyme. Cell Res 27:1341-50. https://doi.org/10.1038/cr.2017.110.\u003c/li\u003e\n\u003cli\u003eCao R, Zeng H, Zhang H (2009) 3D-QSAR studies on a series of inhibitors docked into a new homology model of the DNA-PK receptor. Curr Pharm Des 15:3796-825. https://doi.org/10.2174/138161209789649484.\u003c/li\u003e\n\u003cli\u003eIhmaid S, Al-Rawi J, Bradley C, Angove MJ, Robertson MN, Clark RL (2011) Synthesis, structural elucidation, DNA-PK inhibition, homology modelling and anti-platelet activity of morpholino-substituted-1,3-naphth-oxazines. Bioorg Med Chem 19:3983-94. https://doi.org/10.1016/j.bmc.2011.05.032.\u003c/li\u003e\n\u003cli\u003eClapham KM, Rennison T, Jones G, Craven F, Bardos J, Golding BT, et al. (2012) Potent enantioselective inhibition of DNA-dependent protein kinase (DNA-PK) by atropisomeric chromenone derivatives. Org Biomol Chem 10:6747-57. https://doi.org/10.1039/c2ob26035b.\u003c/li\u003e\n\u003cli\u003eMaira SM, Stauffer F, Brueggen J, Furet P, Schnell C, Fritsch C, et al. (2008) Identification and characterization of NVP-BEZ235, a new orally available dual phosphatidylinositol 3-kinase/mammalian target of rapamycin inhibitor with potent in vivo antitumor activity. Mol Cancer Ther 7:1851-63. https://doi.org/10.1158/1535-7163.MCT-08-0017.\u003c/li\u003e\n\u003cli\u003eMukherjee B, Tomimatsu N, Amancherla K, Camacho CV, Pichamoorthy N, Burma S (2012) The dual PI3K/mTOR inhibitor NVP-BEZ235 is a potent inhibitor of ATM- and DNA-PKCs-mediated DNA damage responses. Neoplasia 14:34-43. https://doi.org/10.1593/neo.111512.\u003c/li\u003e\n\u003cli\u003eHardcastle IR, Cockcroft X, Curtin NJ, El-Murr MD, Leahy JJ, Stockley M, et al. (2005) Discovery of potent chromen-4-one inhibitors of the DNA-dependent protein kinase (DNA-PK) using a small-molecule library approach. J Med Chem 48:7829-46. https://doi.org/10.1021/jm050444b.\u003c/li\u003e\n\u003cli\u003eZhao Y, Thomas HD, Batey MA, Cowell IG, Richardson CJ, Griffin RJ, et al. (2006) Preclinical evaluation of a potent novel DNA-dependent protein kinase inhibitor NU7441. Cancer Res 66:5354-62. https://doi.org/10.1158/0008-5472.CAN-05-4275.\u003c/li\u003e\n\u003cli\u003eCano C, Saravanan K, Bailey C, Bardos J, Curtin NJ, Frigerio M, et al. (2013) 1-substituted (Dibenzo[b,d]thiophen-4-yl)-2-morpholino-4H-chromen-4-ones endowed with dual DNA-PK/PI3-K inhibitory activity. J Med Chem 56:6386-401. https://doi.org/10.1021/jm400915j.\u003c/li\u003e\n\u003cli\u003eMunck JM, Batey MA, Zhao Y, Jenkins H, Richardson CJ, Cano C, et al. (2012) Chemosensitization of cancer cells by KU-0060648, a dual inhibitor of DNA-PK and PI-3K. Mol Cancer Ther 11:1789-98. https://doi.org/10.1158/1535-7163.MCT-11-0535.\u003c/li\u003e\n\u003cli\u003eShinohara ET, Geng L, Tan J, Chen H, Shir Y, Edwards E, et al. (2005) DNA-dependent protein kinase is a molecular target for the development of noncytotoxic radiation-sensitizing drugs. Cancer Res 65:4987-92. https://doi.org/10.1158/0008-5472.CAN-04-4250.\u003c/li\u003e\n\u003cli\u003eFok JHL, Ramos-Montoya A, Vazquez-Chantada M, Wijnhoven PWG, Follia V, James N, et al. (2019) AZD7648 is a potent and selective DNA-PK inhibitor that enhances radiation, chemotherapy and olaparib activity. Nat Commun 10:5065. https://doi.org/10.1038/s41467-019-12836-9.\u003c/li\u003e\n\u003cli\u003eGoldberg FW, Finlay MRV, Ting AKT, Beattie D, Lamont GM, Fallan C, et al. (2020) The Discovery of 7-Methyl-2-[(7-methyl[1,2,4]triazolo[1,5-a]pyridin-6-yl)amino]-9-(tetrahydro-2H-pyran-4-yl)-7,9-dihydro-8H-purin-8-one (AZD7648), a Potent and Selective DNA-Dependent Protein Kinase (DNA-PK) Inhibitor. J Med Chem 63:3461-71. https://doi.org/10.1021/acs.jmedchem.9b01684.\u003c/li\u003e\n\u003cli\u003eWilloughby CE, Jiang Y, Thomas HD, Willmore E, Kyle S, Wittner A, et al. (2020) Selective DNA-PKcs inhibition extends the therapeutic index of localized radiotherapy and chemotherapy. J Clin Invest 130:258-71. https://doi.org/10.1172/JCI127483.\u003c/li\u003e\n\u003cli\u003eZenke FT, Zimmermann A, Sirrenberg C, Dahmen H, Kirkin V, Pehl U, et al. (2020) Pharmacologic Inhibitor of DNA-PK, M3814, Potentiates Radiotherapy and Regresses Human Tumors in Mouse Models. Mol Cancer Ther 19:1091-101. https://doi.org/10.1158/1535-7163.MCT-19-0734.\u003c/li\u003e\n\u003cli\u003eChaplin AK, Hardwick SW, Liang S, Kefala Stavridi A, Hnizda A, Cooper LR, et al. (2021) Dimers of DNA-PK create a stage for DNA double-strand break repair. Nat Struct Mol Biol 28:13-9. https://doi.org/10.1038/s41594-020-00517-x.\u003c/li\u003e\n\u003cli\u003eLiang S, Blundell TL (2023) Human DNA-dependent protein kinase activation mechanism. Nat Struct Mol Biol 30:140-7. https://doi.org/10.1038/s41594-022-00881-w.\u003c/li\u003e\n\u003cli\u003eLiang S, Thomas SE, Chaplin AK, Hardwick SW, Chirgadze DY, Blundell TL (2022) Structural insights into inhibitor regulation of the DNA repair protein DNA-PKcs. Nature 601:643-8. https://doi.org/10.1038/s41586-021-04274-9.\u003c/li\u003e\n\u003cli\u003eHong CR, Liew LP, Wong WW, Dickson BD, Cheng G, Shome A, et al. (2024) Identification of 6-Anilino Imidazo[4,5-c]pyridin-2-ones as Selective DNA-Dependent Protein Kinase Inhibitors and Their Application as Radiosensitizers. J Med Chem 67:12366-85. https://doi.org/10.1021/acs.jmedchem.4c01120.\u003c/li\u003e\n\u003cli\u003eSushko I, Novotarskyi S, Korner R, Pandey AK, Rupp M, Teetz W, et al. (2011) Online chemical modeling environment (OCHEM): web platform for data storage, model development and publishing of chemical information. J Comput Aided Mol Des 25:533-54. https://doi.org/10.1007/s10822-011-9440-2.\u003c/li\u003e\n\u003cli\u003eMauri A. alvaDesc: A Tool to Calculate and Analyze Molecular Descriptors and Fingerprints. In: Roy K, editor. Ecotoxicological QSARs. New York, NY: Springer US; 2020. p. 801-20.\u003c/li\u003e\n\u003cli\u003eMitra S, Halder AK, Koley A, Ghosh N, Panda P, Mandal SC, et al. (2024) Unveiling structural determinants for FXR antagonism in 1,3,4-trisubstituted-Pyrazol amide derivatives: A multi-scale in silico modelling approach. Comput Biol Med 180:108991. https://doi.org/10.1016/j.compbiomed.2024.108991.\u003c/li\u003e\n\u003cli\u003eDanishuddin, Khan AU (2016) Descriptors and their selection methods in QSAR analysis: paradigm for drug design. Drug Discov Today 21:1291-302. https://doi.org/10.1016/j.drudis.2016.06.013.\u003c/li\u003e\n\u003cli\u003eLesani S, Tavalla M, Eslami G, Boozhmehrani MJ (2025) Quantitative Structure-Activity Relationship Modeling and Molecular Docking Studies of TgCDPK1 Inhibitors in Toxoplasma gondii. Microbiologyopen 14:e70039. https://doi.org/10.1002/mbo3.70039.\u003c/li\u003e\n\u003cli\u003eAmbure P, Aher RB, Gajewicz A, Puzyn T, Roy K (2015) \u0026ldquo;NanoBRIDGES\u0026rdquo; software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometrics and Intelligent Laboratory Systems 147:1-13. https://doi.org/https://doi.org/10.1016/j.chemolab.2015.07.007.\u003c/li\u003e\n\u003cli\u003eGramatica P (2007) Principles of QSAR models validation: internal and external. 26:694-701. https://doi.org/https://doi.org/10.1002/qsar.200610151.\u003c/li\u003e\n\u003cli\u003eRoy K, Kar S, Ambure P (2015) On a simple approach for determining applicability domain of QSAR models. Chemometrics and Intelligent Laboratory Systems 145:22-9. https://doi.org/https://doi.org/10.1016/j.chemolab.2015.04.013.\u003c/li\u003e\n\u003cli\u003eTian T, Li S, Fang M, Zhao D, Zeng J (2024) MolSHAP: Interpreting Quantitative Structure-Activity Relationships Using Shapley Values of R-Groups. J Chem Inf Model 64:2236-49. https://doi.org/10.1021/acs.jcim.3c00465.\u003c/li\u003e\n\u003cli\u003eTosco P, Balle T (2011) Open3DQSAR: a new open-source software aimed at high-throughput chemometric analysis of molecular interaction fields. J Mol Model 17:201-8. https://doi.org/10.1007/s00894-010-0684-x.\u003c/li\u003e\n\u003cli\u003eTosco P, Balle T, Shiri F (2011) Open3DALIGN: an open-source software aimed at unsupervised ligand alignment. J Comput Aided Mol Des 25:777-83. https://doi.org/10.1007/s10822-011-9462-9.\u003c/li\u003e\n\u003cli\u003eMondal I, Halder AK, Pattanayak N, Mandal SK, Cordeiro M (2024) Shaping the Future of Obesity Treatment: In Silico Multi-Modeling of IP6K1 Inhibitors for Obesity and Metabolic Dysfunction. Pharmaceuticals (Basel) 17. https://doi.org/10.3390/ph17020263.\u003c/li\u003e\n\u003cli\u003eMitra S, Chatterjee S, Bose S, Panda P, Basak S, Ghosh N, et al. (2024) Finding structural requirements of structurally diverse alpha-glucosidase and alpha-amylase inhibitors through validated and predictive 2D-QSAR and 3D-QSAR analyses. J Mol Graph Model 126:108640. https://doi.org/10.1016/j.jmgm.2023.108640.\u003c/li\u003e\n\u003cli\u003eGhosh A, Panda P, Halder AK, Cordeiro M (2022) In silico characterization of aryl benzoyl hydrazide derivatives as potential inhibitors of RdRp enzyme of H5N1 influenza virus. Front Pharmacol 13:1004255. https://doi.org/10.3389/fphar.2022.1004255.\u003c/li\u003e\n\u003cli\u003eEberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. J Chem Inf Model 61:3891-8. https://doi.org/10.1021/acs.jcim.1c00203.\u003c/li\u003e\n\u003cli\u003eTrott O, Olson AJ (2010) AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31:455-61. https://doi.org/10.1002/jcc.21334.\u003c/li\u003e\n\u003cli\u003eMitra S, Halder AK, Ghosh N, Mandal SC, Cordeiro M (2023) Multi-model in silico characterization of 3-benzamidobenzoic acid derivatives as partial agonists of Farnesoid X receptor in the management of NAFLD. Comput Biol Med 157:106789. https://doi.org/10.1016/j.compbiomed.2023.106789.\u003c/li\u003e\n\u003cli\u003eRoy K, Kar S (2014) The rm2 metrics and regression through origin approach: reliable and useful validation tools for predictive QSAR models (Commentary on \u0026apos;Is regression through origin useful in external validation of QSAR models?\u0026apos;). Eur J Pharm Sci 62:111-4. https://doi.org/10.1016/j.ejps.2014.05.019.\u003c/li\u003e\n\u003cli\u003eTodeschini R, Consonni V (2008) Handbook of molecular descriptors. John Wiley \u0026amp; Sons, Place.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-computer-aided-molecular-design","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jcam","sideBox":"Learn more about [Journal of Computer-Aided Molecular Design](http://link.springer.com/journal/10822)","snPcode":"10822","submissionUrl":"https://submission.nature.com/new-submission/10822/3","title":"Journal of Computer-Aided Molecular Design","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Radiosensitizers, DNA-dependent Protein kinase (DNA-PK), QSAR, molecular dynamics simulation, MolSHAP","lastPublishedDoi":"10.21203/rs.3.rs-7479073/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7479073/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRadiosensitizers are agents that make tumour cells more sensitive to radiation therapy. One key mechanism involves inhibition of the DNA-dependent protein kinase (DNA-PK), an enzyme crucial for repairing DNA double-strand breaks in mammalian cells. Suppression of the DNA-PK enzyme compromises the double-strand break repairs to amplify the radiation induced toxicity among the tumour cells. In this study, 73 6‑Anilino Imidazo[4,5‑c]pyridin-2-one derivatives were curated as potent DNA-PK inhibitors and subjected them to 2D -and 3D-Quantitative Structure Activity Relationship (QSAR) analyses to explore their structural requirements. Apart from conventional methodology, we implemented newly developed MolSHAP analyses for R-group analyses. Significant information regarding structural requirements were retrieved from each of these cheminformatic analyses. Additionally, to understand the interaction between the ligands and the DNA-PK receptor, molecular dynamics (MD) simulation analysis of 100ns were carried out for the most and the least potent compounds among the dataset. The findings indicated H-bond and π-π interactions to be the key factors for binding interactions. Furthermore, novel ligands were designed through the MolSHAP tool and were validated through the chemometric model developed in this investigation. The designed compound exhibited favourable predicted activity and replicated key interaction profiles of the co-crystallized bound ligand in MD simulations. The investigation was carried out through open-access tools to safeguard reproducibility and accessibility among researchers.\u003c/p\u003e","manuscriptTitle":"Multi-scale in-silico modelling to unveil structural requirements for DNA-PK inhibitors as radiosensitizers and MolSHAP based design of novel ligands","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-19 17:33:26","doi":"10.21203/rs.3.rs-7479073/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-23T05:13:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-27T10:18:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"150772791724870258797329423428604728477","date":"2025-09-17T15:45:54+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-14T11:40:37+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-10T02:07:13+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-02T12:43:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Computer-Aided Molecular Design","date":"2025-08-28T10:02:15+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-computer-aided-molecular-design","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jcam","sideBox":"Learn more about [Journal of Computer-Aided Molecular Design](http://link.springer.com/journal/10822)","snPcode":"10822","submissionUrl":"https://submission.nature.com/new-submission/10822/3","title":"Journal of Computer-Aided Molecular Design","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"5796bff1-f9ce-4820-8f5c-ef0ab86ae6fb","owner":[],"postedDate":"September 19th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-12-29T16:02:37+00:00","versionOfRecord":{"articleIdentity":"rs-7479073","link":"https://doi.org/10.1007/s10822-025-00733-x","journal":{"identity":"journal-of-computer-aided-molecular-design","isVorOnly":false,"title":"Journal of Computer-Aided Molecular Design"},"publishedOn":"2025-12-22 15:57:15","publishedOnDateReadable":"December 22nd, 2025"},"versionCreatedAt":"2025-09-19 17:33:26","video":"","vorDoi":"10.1007/s10822-025-00733-x","vorDoiUrl":"https://doi.org/10.1007/s10822-025-00733-x","workflowStages":[]},"version":"v1","identity":"rs-7479073","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7479073","identity":"rs-7479073","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}