Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening

Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening View ORCID Profile Jingyu Yang , Meng Wang , Jürgen Dönitz , Björn Chapuy , View ORCID Profile Tim Beißbarth doi: https://doi.org/10.1101/2024.06.20.24309094 Jingyu Yang 1 Medical Bioinformatics, University Medical Center Göttingen , Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jingyu Yang For correspondence: tim.beissbarth{at}bioinf.med.uni-goettingen.de Meng Wang 2 Department of Hematology, Oncology, and Cancer Immunology, Charité -University Medical Center Berlin , Campus Benjamin Franklin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jürgen Dönitz 1 Medical Bioinformatics, University Medical Center Göttingen , Germany 3 Campus-Institute Data Science (CIDAS), University of Göttingen , Göttingen, Germany 4 Comprehensive Cancer Center Niedersachsen (CCCN) , Göttingen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Björn Chapuy 2 Department of Hematology, Oncology, and Cancer Immunology, Charité -University Medical Center Berlin , Campus Benjamin Franklin, Germany 5 Department of Hematology and Medical Oncology, Georg-August University Göttingen , Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Tim Beißbarth 1 Medical Bioinformatics, University Medical Center Göttingen , Germany 3 Campus-Institute Data Science (CIDAS), University of Göttingen , Göttingen, Germany 4 Comprehensive Cancer Center Niedersachsen (CCCN) , Göttingen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tim Beißbarth For correspondence: tim.beissbarth{at}bioinf.med.uni-goettingen.de Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Identifying and validating genotype-guided drug combinations for a specific molecular subtype in cancer therapy represents an unmet medical need and is important in enhancing efficacy and reducing toxicity. However, the exponential increase in combinatorial possibilities constrains the ability to identify and validate effective drug combinations. In this context, we have developed Onko DrugCombScreen , an innovative tool aiming at advancing precision medicine based on identifying significant drug combination candidates in a target cancer cohort compared to a comparison cohort. Onko DrugCombScreen , inspired by the Molecular Tumor Board (MTB) process, synergizes drug knowledge-base analysis with various statistical methodologies and data visualization techniques to pinpoint drug combination candidates. Validated through a TCGA-BRCA case study, Onko DrugCombScreen has demonstrated its proficiency in discerning established drug combinations in a specific cancer type and in revealing potential novel drug combinations. By enhancing the capability of drug combination discovery through drug knowledge bases, Onko DrugCombScreen represents a significant advancement in personalized cancer treatment by identifying promising drug combinations, setting the stage for the development of more precise and potent combination treatments in cancer care. The Onko DrugCombScreen shiny app is available at https://rshiny.gwdg.de/apps/onko_drugcombscreen/ . The Git repository can be accessed at https://gitlab.gwdg.de/MedBioinf/mtb/onko_drugcombscreen . Introduction Cancer treatment is an intricate field, with the ongoing quest to develop therapies that effectively target the disease while minimizing side effects. The application of synergistic drug combinations builds on the idea of lowering the concentration of both drugs to achieve the same effect with fewer side effects and represents an important advancement in cancer therapy [ 9 ]. This approach aims to overcome the limitations of single-agent treatments by improving efficacy and reducing the likelihood of drug resistance [ 7 ]. However, the task of identifying optimal drug combinations is complicated by the significant variability in tumor types and patient responses, along with the complexities of cancer biology, the high dimensionality of data, and the number of drug combinations far beyond what is possible for clinical testing [ 18 , 23 ]. These challenges make it difficult to predict which combinations are most likely to be effective in specific cancer molecular type cohorts compared to other molecular subtypes, necessitating advanced computational tools and extensive experimental validation to navigate the vast landscape of potential drug combinations and tailor treatments to cohort patients’ needs [ 27 ]. Molecular Tumor Boards (MTBs) are crucial in personalizing cancer treatment, integrating multidisciplinary expertise to interpret genetic data and guide treatment decisions based on a patient’s unique tumor characteristics [ 25 , 21 , 19 ]. Drawing inspiration from the methodologies employed by MTBs, the utilization of drug databases alongside detailed drug’s level of evidence information emerges as a crucial strategy in advancing patient-specific treatment recommendations [ 25 , 21 , 19 ]. This approach not only facilitates the identification of the most suitable drugs for individual patients based on their genetic profiles but also sets the foundation for drug combination prediction based on patient cohorts. Applying statistical methods based on drug databases to the set of recommended drugs for these patient cohorts enables researchers and clinicians to predict more accurately effective drug combinations. This strategy underscores a significant shift towards a data-driven and evidence-based framework to optimize combination therapy for cancer patients, leveraging the increasingly available genetic and pharmacological data to enhance treatment efficacy and patient outcomes. In recent decades, an increasing application of computational approaches has been developed for the prediction of drug combinations and their effects. Preuer et al. developed DeepSynergy , a deep learning-based approach that accurately predicts drug combination synergies for cancer treatments, significantly surpassing traditional performance methods [ 26 ]. Similarly, Wang et al. introduced DeepDDS , a deep learning model that employs graph neural networks and attention mechanisms to precisely predict and prioritize synergistic drug combinations for cancer treatments, achieving the advantage of enhanced interpretability through chemical substructure analysis [ 33 ]. Cheng et al. demonstrated that a network-based methodology, concentrating on the relative configuration of drug–target modules in connection to disease modules, can effectively prioritize potentially efficacious drug combinations for complex diseases such as cancer [ 5 ]. GAECDS , presented by Li et al., is an innovative approach combining graph autoencoders and convolutional neural networks to accurately predict drug synergy, showing superior performance in identifying efficacious drug combinations [ 20 ]. Concurrently, numerous classical machine learning models have also exhibited performance comparable to deep learning methods, demonstrating their robustness and utility in this complex domain. Gayvert et al. showcased that a Random Forest model, utilizing single drug dose responses as features, could accurately predict drug pair synergy and effectiveness in mutant BRAF melanomas [ 11 ]. Janizek et al. introduced TreeCombo , an XGBoost-based approach that leverages the power of gradient boosting to improve predictive accuracy, outperforming DeepSynergy by using drug physiochemical features and cancer cell line gene expression data. The use of XGBoost, which combines multiple decision trees to make robust predictions, demonstrated comparable efficacy to deep learning on medium-scale datasets, while offering the additional benefits of reduced complexity in hyperparameter tuning and enhanced interpretability through TreeSHAP , a feature attribution method that identifies the contribution of each variable in a clear and consistent manner [ 17 ]. However, current pre-clinical screenings primarily focus on the synergistic effects of drug combinations, often overlooking key factors for clinical success such as potential toxicity and selective efficacy against tumors [ 18 ]. At the same time, there is a clear lack of innovative computational solutions to demonstrate their feasibility and benefits in translational applications, especially in the field of cancer, where there is an urgent need to identify combination therapies suitable for specific cancer group patients based on patient-specific biomarkers [ 6 , 29 ]. In this paper, we present an Onko DrugCombScreen Shinyapp designed to address this gap in cancer therapy which could predict the significant drug combination candidates based on the target patient cohort statistical analysis against the comparison cohort. The primary goal of Onko DrugCombScreen extends beyond merely providing treatment recommendations based on drug databases like GDKD [ 8 ], CiVIC [ 13 ], and OncoKB [ 4 ]. It integrates statistical methods and data visualization to analyze target cancer type cohort and comparison cohort genetic data against extensive drug databases, thereby uncovering potential drug combinations and mapping them onto cell line data, providing a robust basis for clinical drug screening. Based on the drug evidence levels in the knowledge database for medications, one can directly ascertain whether the variant mapping drugs are selective at the cancer types in the target patient cohort, and the previous studies collected in the database can save workload on drug toxicity analysis. This brings renewed hope for the clinical translation of cancer-type-specific drug combination therapies. View this table: View inline View popup Download powerpoint Table 1. Contingency table for Fisher exact test analysis. In this table, a represents the number of patients in the target tumor cohort receiving Drug1+Drug2 co-recommendation, b is the number of comparison cohort patients receiving Drug1+Drug2 co-recommendation, c is the number of patients in the target tumor cohort not receiving Drug1+Drug2 co-recommendation, and d is the number of comparison cohort patients not receiving Drug1+Drug2 co-recommendation. Methods Fisher’s exact test in subtype recommendations Besides predicting single drugs, clinicians and researchers are interested in determining whether two drugs are simultaneously recommended for the target tumor type and exhibit significant differences compared to the comparison tumor group. Here, we defined co-recommended drugs as candidate drug combinations that are presented in the Drug_comb column in the DrugComb analysis table. We then counted the number of patients in the target tumor cohort in the comparison cohort for each candidate drug combination. Subsequently, we used these four counts to construct a contingency table (Table ?? ) and performed a Fisher’s exact test for each candidate drug combination. By analyzing the p-value and odds ratio that are circled with a red rectangle in Figure 1 results obtained from Fisher’s exact test, we can determine if the occurrence of a candidate drug combination is significantly different and assess the magnitude of this difference. The p-values were adjusted using the Benjamini-Hochberg method, as reflected in the adjust_p.value column, to account for multiple hypothesis testing and control the false discovery rate (FDR). Additionally, we report drug combination candidate recommendations for cell line data in the final four columns in the DrugComb analysis table to assist with wet-lab validation. ( Fig. 1 ). Download figure Open in new tab Fig. 1. A ‘drug co-recommendation’ refers to the recommendation of two drugs for a single patient, exemplified as Drug1 + Drug2. The bar plot illustrates the counts of such co-recommendations within the target cancer subtype cohort compared to the comparison cancer subtype cohort, leading to the construction of the contingency table on the right. Fisher’s exact test was applied to each drug co-recommendation to assess statistical significance, resulting in the DrugComb analysis table displayed below. The percentage column ( percentage ) indicates the proportion of drug co-recommendations, with the p-value and adjusted p-value columns ( adjust_p.value , adjust_p.value ) reflecting the significance level. The odds ratio ( oddsRatio ) provides a measure of the effect size in comparison to the control group. The final four columns offer details on the cell line drug recommendation status, including specific cell line IDs ( individual_id ) used for wet lab validation. Here, Fisher’s exact test serves as a robust statistical method to determine the significance of the association between the candidate drug combination and the tumor type. Fisher’s exact test is particularly suitable for small sample sizes and for datasets where the assumptions of chi-squared tests are not met. The null hypothesis for Fisher’s exact test posits that there is no association between the drug combination and the tumor type. Under this assumption, the test calculates the p-value, which represents the probability of observing the current distribution of patients, or a more extreme distribution, if the null hypothesis were true. The p-value is computed using the hypergeometric distribution formula: Here, a, b, c , and d are the cell counts in the 2 × 2 table, and n is the total sample size ( a + b + c + d ). The p -value is then the sum of probabilities ( P ) for all arrangements of the table where the association between the rows and columns is as extreme as or more extreme than in the observed table. A low p -value indicates that the observed distribution is unlikely under the null hypothesis, suggesting a significant association between candidate drug combinations and the tumor type. The odds ratio is calculated as , representing the odds of the target tumor group receiving Drug1+Drug2 compared to the odds of the comparison group receiving the same. An odds ratio greater than 1 indicates a higher likelihood of the target cohort receiving the drug combination compared to the comparison cohort, while an odds ratio less than 1 suggests the opposite. Drug Level of Evidence Here, we adopted the MTB drug’s level of evidence category approach proposed by Perera-Bel [ 25 ]. As shown in Table 2 . “ A ” signifies evidence for the same cancer type, while “ B “ indicates evidence for any other cancer type. Horizontally, Level 1 represents evidence supported by regulatory agencies or clinical guidelines. Level 2 includes evidence from clinical trials. Finally, Level 3 consists of preclinical trial evidence. Therefore, based on the different target cancer types of drugs and their respective clinical evidence, six levels of drug evidence are derived: A1, A2, A3, B1, B2, B3. With this drug level of evidence, the selection of recommended drugs for specific cancer types and their clinical strength can be clearly defined, which can guide the clinical decision. View this table: View inline View popup Download powerpoint Table 2. Drugs obtained from the Drug knowledge database are classified into clinically relevant categories using a system of six levels of evidence. Material Analysis tools Onko DrugCombScreen was implemented using R (v.4.3.1) and R Shiny (v.1.8.0). This shiny app integrated a variety of R programming language packages for comprehensive bioinformatics analysis. For parsing and generating data structures, we utilized readxl v1.4.1 [ 38 , 37 ]. To facilitate data manipulation and transformation, we employed packages like reshape2 v1.4.4 [ 36 ], tidyr v1.2.1 [ 37 ], and dplyr v1.0.10 [ 39 ]. We applied packages such as maftools v2.12.0 [ 22 ], clusterProfiler v4.4.4 [ 41 ], and VariantAnnotation v1.42.1 [ 24 ] for the analysis of somatic variants, functional profiles of genes. For data visualization, we used packages such as circlize v0.4.15 [ 15 ] for circular visualizations, ggalluvial v0.12.3 [ 3 , 2 ] for alluvial diagrams, ggrepel v0.9.2 [ 28 ] for label clarity, ComplexHeatmap v2.12.1 [ 14 ] for sophisticated heatmaps, and ggplot2 v3.4.0 [ 31 ] for creating customizable static plots. Data Source The harmonized drug database, is derived from open-source drug knowledge databases including GDKD [ 8 ], CiVIC [ 13 ], and OncoKB [ 4 ], utilizes the DrugBank Vocabulary dataset from DrugBank [ 40 ] to standardize drug synonyms. TCGA-BRCA data and breast cancer cell line data used in the case study were collected from UCSC Xena hubs [ 12 ]. Results Case Study: Application and Validation Using TCGA-BRCA Data Dataset Selection and Processing In this case study, the Onko DrugCombScreen was applied to the TCGA-BRCA dataset to validate its efficacy in identifying effective drug combinations for breast cancer. The TCGA-BRCA dataset, derived from the TCGA Pan-Cancer (PANCAN) initiative, was chosen for its comprehensive genetic profiling, including extensive data on copy number variations (CNVs), single nucleotide variations (SNVs), and molecular subtype profiles [ 35 ]. This dataset provides a broad coverage of genetic variations, making it an ideal resource for this analysis. Additionally, cell line data from the Cancer Cell Line Encyclopedia (Breast) were incorporated to complement the analysis [ 1 ]. All of the above data set are available in UCSC Xena hubs [ 12 ]. In the preprocessing phase, somatic mutation data from the TCGA Pan-Cancer (PANCAN) was converted into a compatible CSV file for analysis by the Onko DrugCombScreen . This process involved filtering the dataset to isolate BRCA cancer data and further stratifying it into molecular subtypes - Luminal A/B, HER2+, Normal-like, and Basal-like. In this case study, we used Normal-like breast cancer as the comparison cohort, and HER2+, and Basal-like subtypes as the target cohorts, respectively, to analyze and validate the efficacy of Onko DrugCombScreen . Additionally, by integrating cell line data, Onko DrugCombScreen provided guidance on suitable cell lines for subsequent experimental validation. Validation And Results The drug co-recommendation comparison analysis revealed significant disparities between the three BRCA subtypes (HER2+ and Basal-like) and the Normal-like BRCA data. Significant drug co-recommendations extracted from Onko DrugCombScreen were compared with combinational therapies in Wang’s review [ 34 ], as well as FDA-approved drug combinations to validate the effectiveness. As the supplementary table A shows, the “adjust p.value” and “OR” (odds ratio), obtained from the Fisher’s exact test, indicate the significance and magnitude of the drug combination in the target cohort compared to the comparison cohort, the “Percentage” depicts the proportion of the drug combination recommended in the target cohort. Setting the threshold at adjust p . value 1, and Percentage > 50% retains around 30% of the significant candidate drug combinations (30250/111987 in HER2+ vs Normal-like and 48348/112069 in Basal-like vs Normal-like). Notably, these stringent criteria preserved almost all approved and clinical trial drug combinations, including the approved combinational therapy of Pertuzumab + Trastuzumab for the HER2+ subtype and Pembrolizumab + Paclitaxel for Triple-Negative Breast Cancer (TNBC). These results highlight Onko DrugCombScreen ’s accuracy in identifying clinically relevant drug combinations, confirming its effectiveness. Besides, upon comparison with the DrugComb.org database [ 42 ], it was found that none of the approved and currently in clinical trial drug combinations of breast cancer had any recorded synergy scores. The validation analysis demonstrates that the Onko DrugCombScreen is adept at identifying established breast cancer drug combinations in the BRCA subtypes like HER2+ and Basal-like when compared to Normal-like BRCA subtype. This finding not only validates the tool’s effectiveness but also highlights its potential in discovering novel drug combinations for various cancer types. Consequently, the case study accentuates the utility of the Onko DrugCombScreen in providing targeted and efficacious drug recommendations. Data Analysis Workflow of Onko DrugCombScreen The data analysis workflow of Onko DrugCombScreen is depicted in Figure 2 : Variant data such as single nucleotide variants (SNV) and copy number variants (CNV) from both the cancer subtype cohort and the comparison cancer subtype cohort are preprocessed and converted into variant tables compatible with Onko DrugCombScreen . These patient variant data are then mapped to public drug databases (CiVIC, GDKD, OncoKb) after integration with variant interpretation annotations and drug evidence levels for drug recommendations. The resulting drug recommendations are subjected to statistical analysis, focusing on the statistical differences in drug combination candidates observed between the target cancer subtype group and the comparison cancer subtype group. To identify drug combination candidates that are significantly and frequently recommended in the target group compared to the comparison group, Fisher’s exact test is applied. Subsequently, the selected drug combination candidates undergo an integrated analysis with cell line data to identify available cell line samples, facilitating wet-lab validation. Additionally, all analysis results are visualized, making the findings clearer and more intuitive. The integration of these processes is crucial for confirming drug combination recommendations for the cancer type of interest. The final validation stage may include conducting wet-lab drug screenings to confirm the analysis results and deepen the understanding of the underlying biological mechanisms ( Fig. 2 ). Download figure Open in new tab Fig. 2. The workflow for Onko DrugCombScreen drug combination data analysis. After the recommendation process based on drug knowledge, SNVs and CNVs are merged into an annotated drug table. Following statistical analysis and integration of cell line data, the final DrugComb analysis table will be used for visualization and wet lab validation. Data Preprocessing SNVs and CNVs are typically stored in formats such as VCF, MAF, TXT, or Excel. A preprocessing step is necessary to convert these various formats into csv format ( Fig 2 ). These dataframes are then suitable for use in the knowledge-based drug recommendation analysis within Onko DrugCombScreen . Matching Rule Between Variant Annotations in Patients’ data and Database Due to the different annotation descriptions of variants in the three-drug databases (GDKD [ 8 ], CiVIC [ 13 ], and OncoKB [ 4 ]) and original patients’ variants data, we harmonized the three drug databases and designed a matching rule based on the interpretation of biological significance ( Table 3 ). All variant classes or effects map to the biological interpretations of “loss”, “gain”, or “mutation”. We can then associate the original variants ( Table 4 ) with the information in the knowledge database based on biological interpretations and obtain the relevant target drug information. View this table: View inline View popup Download powerpoint Table 3. Biological interpretation of variant annotations in drug databases. View this table: View inline View popup Download powerpoint Table 4. Biological interpretation of variant annotations in patients’ variants. Drug Recommendation Annotation After the matching rule is defined, the drug knowledge-based analysis was performed to export the drug recommendation tables across all target and comparison subtype data. However, due to discrepancies in drug nomenclature across the three drug databases, we employed the “DrugBank Vocabulary” dataset from DrugBank to standardize synonymous drug names. Subsequently, each drug name was annotated to its final drug class. This annotation is stored in the columns “Origin Drug Name” and “Classified Drug Name” of the DrugComb analysis table. Additionally, other useful information such as variant type is annotated in the “mutType” column, and variant match status—which indicates whether the amino acid change in the raw data exactly matches the database records or not—is saved in the “Match Sign” column. Data visualization Onko DrugCombScreen provides a variety of charts for visual analysis results, allowing users to understand data more intuitively ( Fig. 3 ). The application integrates multiple plotting functions, including volcano plots, heatmaps, circle plots, alluvial diagrams, upset plots, and bar charts. These visualization tools help to easily identify recommended drugs or candidate drug combinations for subsequent wet-lab analysis and validation. Users can configure settings in the left panel of Onko DrugCombScreen and customize the resolution for PDF file export. Download figure Open in new tab Fig. 3. Visualization of the Onko DrugCombScreen . (A) Volcano plot identifying significant drug combinations. (B) Circle plot depicting the most proportional drug co-recommendations. (C) Alluvial diagram tracing mutations back to recommended single drugs. (D) UpSet plot showing the top-recommended single drugs and their intersections. Discussion Drug combinations are widely recognized for their benefits in cancer therapy. Here, we developed the Onko DrugCombScreen shiny app integrated statistical analysis to identify the most significant candidate drug combinations for a target tumor type cohort. We utilized drug knowledge base recommendations derived from mutation data of the targeted cancer patient cohort and the comparison cohort to identify drug co-recommendations. This is complemented by integrating cell line data to assist in the validation of biological experiments. Onko DrugCombScreen ’s ability to identify effective drug combinations, as demonstrated in the TCGA-BRCA case study, suggests a promising way toward more tumor-type tailored and effective cancer combination therapy. In contrast to current computational methods that mainly focus on synergy and dose-response matrices, Onko DrugCombScreen is a drug knowledge-based analysis approach. It not only provides therapeutic recommendations but also offers guidance for clinical research, thereby integrating more closely with clinical applications. Moreover, all drug recommendations can be traced back to patient genetics and variants through the visual alluvial diagram of Onko DrugCombScreen . Utilizing the drug database which is also employed by the MTB report, each recommended drug’s level of evidence and response is explicitly defined. This clarity effectively aids in addressing issues of selectivity in the recommended drug combinations, issues that are often overlooked in previous methods. Moreover, the utilization of drug databases for the recommendation of candidate drug combinations based on patient gene mutation profiles can potentially reduce the effort required for toxicity analysis [ 32 , 10 ]. These databases provide valuable information on the relationships between individual drugs and specific gene mutations or molecular targets, which can guide the selection of drug combinations with potentially favorable efficacy profiles. Furthermore, the drugs included in these databases are often approved or under clinical trials, meaning that their toxicity profiles have been extensively studied and characterized [ 16 , 30 ]. This existing safety data can serve as a foundation for assessing the toxicity of drug combinations, as it provides insights into the common adverse events, dose-limiting toxicities, and recommended dosing schedules of the individual drugs. By leveraging this information, researchers can streamline the toxicity assessment process and make more informed decisions when designing drug combination studies. However, it is crucial to acknowledge that the toxicity of a drug combination may not be a simple sum of the individual drug toxicities and thorough safety assessments of the specific combination, considering factors such as drug-drug interactions, dosing, scheduling, and special patient populations, are still necessary to ensure a comprehensive understanding of the toxicity profile. Looking forward, the potential for further development of the Onko DrugCombScreen is substantial. Due to the lack of data on synergy and dose specificity in the drug database, Onko DrugCombScreen is currently unable to provide such information. In future iterations, it is feasible to incorporate synergy and dose specificity data. Moreover, advancements in artificial intelligence and machine learning can be leveraged to enhance data analysis based on drug knowledge, thereby continuously improving the predictive accuracy and relevance of the tool. Conclusion In conclusion, the Onko DrugCombScreen Shiny app represents an innovative tool in the field of precision cancer therapy, offering a novel drug knowledge-based approach to drug combination screening. Harnessing the power of drug knowledge database analysis, this application integrates advanced statistical analysis and data visualization techniques to explore and identify effective drug combinations. It effectively utilizes drug recommendations from targeted cancer cohort and comparison cohort, combined with cell line data, to provide prominent drug co-recommendations for targeted cancer type. Validated through a TCGA-BRCA case study, the application has demonstrated its potential in accurately identifying both existing and novel drug combinations, aligning with the evolving field of precision oncology. Looking ahead, the integration of artificial intelligence and machine learning technologies holds the promise of further enhancing its predictive capabilities, making it a valuable tool in the quest for more targeted and effective cancer treatment approaches. Data Availability All data produced are available online at TCGA repository Supporting information Supplementary File S1: Sample data files for Onko DrugCombScreen testing This file includes three CSV files: Primary_Her2_mutationSNVCNV DF.csv : Target cancer subtype data. Comparison_Normallike_BRCA_mutation DF.csv : comparison cancer subtype data. Cellline_CCLEBRCA_Mutation_DF.csv : Cell line data. Supplementary Table S1. Comparison of Current Approved Therapies and Clinical Trial Combinations for Breast Cancer Treatment with Onko DrugCombScreen Based Screening result [ 34 ] . Competing interests No competing interest is declared. Funding This project was supported by the Volkswagen Foundation within research project MTB-Report (ZN3424). Author contributions statement JY, MW, BC, JD, and TB designed the study. JY and BC provided major contributions to the computational method and visualization. JY designed, developed, and implemented the Shiny app. MW tested the Shiny app and helped identify problems and bugs. JY wrote the manuscript. All authors critically reviewed the content and approved the final manuscript. Acknowledgments The authors would like to acknowledge the Volkswagen Foundation’s support in MTB-report project. JY was supported by the Ph.D. program “Genome Science” - International Max Plank Research School, part of the Göttingen Graduate Center for Neurosciences, Biophysics, and Molecular Biosciences. JD and TB are members of the Göttingen Campus Institute Data Science. Footnotes Revise the DrugComb plot, and revise the supplementary table. References 1. ↵ Jordi Barretina , Giordano Caponigro , Nicolas Stransky , Kavitha Venkatesan , Adam A Margolin , Sungjoon Kim , Christopher J Wilson , Joseph Lehár , Gregory V Kryukov , Dmitriy Sonkin , et al. The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity . Nature , 483 ( 7391 ): 603 – 607 , 2012 . OpenUrl CrossRef PubMed Web of Science 2. ↵ Jason Cory Brunson . ggalluvial: Layered grammar for alluvial plots . Journal of Open Source Software , 5 ( 49 ): 2017 , 2020 . OpenUrl 3. ↵ Jason Cory Brunson and Quentin D. Read . ggalluvial: Alluvial plots in ‘ggplot2’ , 2023 . R package version 0.12.5. 4. ↵ Debyani Chakravarty , Jianjiong Gao , Sarah Phillips , Ritika Kundra , Hongxin Zhang , Jiaojiao Wang , Julia E Rudolph , Rona Yaeger , Tara Soumerai , Moriah H Nissan , et al. Oncokb: a precision oncology knowledge base . JCO precision oncology , 1 : 1 – 16 , 2017 . OpenUrl CrossRef 5. ↵ Feixiong Cheng , István A Kovács , and Albert-László Barabási . Network-based prediction of drug combinations . Nature communications , 10 ( 1 ): 1197 , 2019 . OpenUrl 6. ↵ Wenqiang Cui , Adnane Aouidate , Shouguo Wang , Qiuliyang Yu , Yanhua Li , and Shuguang Yuan . Discovering anti-cancer drugs via computational methods . Frontiers in pharmacology , 11 : 733 , 2020 . OpenUrl 7. ↵ João MA Delou , Alana SO Souza , Leonel CM Souza , and Helena L Borges . Highlights in resistance mechanism pathways for combination therapy . Cells , 8 ( 9 ): 1013 , 2019 . OpenUrl 8. ↵ Rodrigo Dienstmann , In Sock Jang , Brian Bot , Stephen Friend , and Justin Guinney . Database of genomic biomarkers for cancer drugs and clinical targetability in solid tumors . Cancer discovery , 5 ( 2 ): 118 – 123 , 2015 . OpenUrl Abstract / FREE Full Text 9. ↵ Diana Duarte and Nuno Vale . Evaluation of synergism in drug combinations and reference models for future orientations in oncology . Current Research in Pharmacology and Drug Discovery , 3 : 100110 , 2022 . OpenUrl 10. ↵ Cristiano Galletti , Patricia Mirela Bota , Baldo Oliva , and Narcis Fernandez-Fuentes . Mining drug–target and drug– adverse drug reaction databases to identify target–adverse drug reaction relationships . Database , 2021 : baab068 , 2021 . OpenUrl 11. ↵ Kaitlyn M Gayvert , Omar Aly , James Platt , Marcus W Bosenberg , David F Stern , and Olivier Elemento . A computational approach for identifying synergistic drug combinations . PLoS computational biology , 13 ( 1 ): e1005308 , 2017 . OpenUrl 12. ↵ Mary J Goldman , Brian Craft , Mim Hastie , Kristupas Repečka , Fran McDade , Akhil Kamath , Ayan Banerjee , Yunhai Luo , Dave Rogers , Angela N Brooks , et al. Visualizing and interpreting cancer genomics data via the xena platform . Nature biotechnology , 38 ( 6 ): 675 – 678 , 2020 . OpenUrl CrossRef PubMed 13. ↵ Malachi Griffith , Nicholas C Spies , Kilannin Krysiak , Joshua F McMichael , Adam C Coffman , Arpad M Danos , Benjamin J Ainscough , Cody A Ramirez , Damian T Rieke , Lynzey Kujan , et al. Civic is a community knowledgebase for expert crowdsourcing the clinical interpretation of variants in cancer . Nature genetics , 49 ( 2 ): 170 – 174 , 2017 . OpenUrl CrossRef PubMed 14. ↵ Zuguang Gu , Roland Eils , and Matthias Schlesner . Complex heatmaps reveal patterns and correlations in multidimensional genomic data . Bioinformatics , 32 ( 18 ): 2847 – 2849 , 2016 . OpenUrl CrossRef PubMed 15. ↵ Zuguang Gu , Lei Gu , Roland Eils , Matthias Schlesner , and Benedikt Brors . “ circlize” implements and enhances circular visualization in r . 2014 . 16. ↵ F Peter Guengerich . Mechanisms of drug toxicity and relevance to pharmaceutical development . Drug metabolism and pharmacokinetics , 26 ( 1 ): 3 – 14 , 2011 . OpenUrl CrossRef PubMed 17. ↵ Joseph D Janizek , Safiye Celik , and Su-In Lee . Explainable machine learning prediction of synergistic drug combinations for precision cancer medicine . BioRxiv , page 331769 , 2018 . 18. ↵ Weikaixin Kong , Gianmarco Midena , Yingjia Chen , Paschalis Athanasiadis , Tianduanyi Wang , Juho Rousu , Liye He , and Tero Aittokallio . Systematic review of computational methods for drug combination prediction . Computational and structural biotechnology journal , 20 : 2807 – 2814 , 2022 . OpenUrl 19. ↵ Nadine S Kurz , Júlia Perera-Bel , Charlotte Höltermann , Tim Tucholski , Jingyu Yang , Tim Beissbarth , and Jürgen Dönitz . Identifying actionable variants in cancer-the dual web and batch processing tool mtb-report . In GMDS , pages 73 – 80 , 2022 . 20. ↵ Huijun Li , Lin Zou , Jamal AH Kowah , Dongqiong He , Lisheng Wang , Mingqing Yuan , and Xu Liu . Predicting drug synergy and discovering new drug combinations based on a graph autoencoder and convolutional neural network . Interdisciplinary Sciences: Computational Life Sciences , 15 ( 2 ): 316 – 330 , 2023 . OpenUrl 21. ↵ Claudio Luchini , Rita T Lawlor , Michele Milella , and Aldo Scarpa . Molecular tumor boards in clinical practice . Trends in cancer , 6 ( 9 ): 738 – 744 , 2020 . OpenUrl 22. ↵ Anand Mayakonda , De-Chen Lin , Yassen Assenov , Christoph Plass , and H Phillip Koeffler . Maftools: efficient and comprehensive analysis of somatic variants in cancer . Genome research , 28 ( 11 ): 1747 – 1756 , 2018 . OpenUrl Abstract / FREE Full Text 23. ↵ Ravi S Narayan , Piet Molenaar , Jian Teng , Fleur MG Cornelissen , Irene Roelofs , Renee Menezes , Rogier Dik , Tonny Lagerweij , Yoran Broersma , Naomi Petersen , et al. A cancer drug atlas enables synergistic targeting of independent drug vulnerabilities . Nature communications , 11 ( 1 ): 2935 , 2020 . OpenUrl 24. ↵ Valerie Obenchain , Michael Lawrence , Vincent Carey , Stephanie Gogarten , Paul Shannon , and Martin Morgan . Variantannotation: a bioconductor package for exploration and annotation of genetic variants . Bioinformatics , 30 ( 14 ): 2076 – 2078 , 2014 . OpenUrl CrossRef PubMed Web of Science 25. ↵ Júlia Perera-Bel , Barbara Hutter , Christoph Heining , Annalen Bleckmann , Martina Fröhlich , Stefan Fröhling , Hanno Glimm , Benedikt Brors , and Tim Beißbarth . From somatic variants towards precision oncology: evidence-driven reporting of treatment options in molecular tumor boards . Genome medicine , 10 : 1 – 15 , 2018 . OpenUrl 26. ↵ Kristina Preuer , Richard PI Lewis , Sepp Hochreiter , Andreas Bender , Krishna C Bulusu , and Günter Klambauer . Deepsynergy: predicting anti-cancer drug synergy with deep learning . Bioinformatics , 34 ( 9 ): 1538 – 1546 , 2018 . OpenUrl 27. ↵ Deepraj Sarmah , Wesley O Meredith , Ian K Weber , Madison R Price , and Marc R Birtwistle . Predicting anti-cancer drug combination responses with a temporal cell state network model . PLoS computational biology , 19 ( 5 ): e1011082 , 2023 . OpenUrl 28. ↵ Kamil Slowikowski . ggrepel: Automatically Position Non-Overlapping Text Labels with ‘ggplot2’ , 2024 . https://ggrepel.slowkow.com/ , https://github.com/slowkow/ggrepel . 29. ↵ Aaron C Tan , Stephen J Bagley , Patrick Y Wen , Michael Lim , Michael Platten , Howard Colman , David M Ashley , Wolfgang Wick , Susan M Chang , Evanthia Galanis , et al. Systematic review of combinations of targeted or immunotherapy in advanced solid tumors . Journal for immunotherapy of cancer , 9 ( 7 ), 2021 . 30. ↵ Andrey A Toropov , Alla P Toropova , Ivan Raska Jr . , Danuta Leszczynska , and Jerzy Leszczynski . Comprehension of drug toxicity: software and databases . Computers in biology and medicine , 45 : 20 – 25 , 2014 . OpenUrl 31. ↵ Randle Aaron M Villanueva and Zhuo Job Chen . ggplot2: elegant graphics for data analysis , 2019 . 32. ↵ Andy H Vo , Terry R Van Vleet , Rishi R Gupta , Michael J Liguori , and Mohan S Rao . An overview of machine learning and big data for drug toxicity evaluation . Chemical research in toxicology , 33 ( 1 ): 20 – 37 , 2019 . OpenUrl 33. ↵ Jinxian Wang , Xuejun Liu , Siyuan Shen , Lei Deng , and Hui Liu . Deepdds: deep graph neural network with attention mechanism to predict synergistic drug combinations . Briefings in Bioinformatics , 23 ( 1 ): bbab390 , 2022 . OpenUrl 34. ↵ Yiling Wang and Audrey Minden . Current molecular combination therapies used for the treatment of breast cancer . International journal of molecular sciences , 23 ( 19 ): 11046 , 2022 . OpenUrl 35. ↵ John N Weinstein , Eric A Collisson , Gordon B Mills , Kenna R Shaw , Brad A Ozenberger , Kyle Ellrott , Ilya Shmulevich , Chris Sander , and Joshua M Stuart . The cancer genome atlas pan-cancer analysis project . Nature genetics , 45 ( 10 ): 1113 – 1120 , 2013 . OpenUrl CrossRef PubMed 36. ↵ Hadley Wickham . Reshaping data with the reshape package . Journal of statistical software , 21 : 1 – 20 , 2007 . OpenUrl 37. ↵ Hadley Wickham , Mara Averick , Jennifer Bryan , Winston Chang , Lucy D’Agostino McGowan , Romain François , Garrett Grolemund , Alex Hayes , Lionel Henry , Jim Hester , et al. Welcome to the tidyverse . Journal of open source software , 4 ( 43 ): 1686 , 2019 . OpenUrl CrossRef 38. ↵ Hadley Wickham and Jennifer Bryan . readxl: Read Excel Files , 2023 . https://readxl.tidyverse.org , https://github.com/tidyverse/readxl . 39. ↵ Hadley Wickham , Romain François , Lionel Henry , Kirill Müller , et al. dplyr: A grammar of data manipulation . 2020 . 40. ↵ David S Wishart , Yannick D Feunang , An C Guo , Elvis J Lo , Ana Marcu , Jason R Grant , Tanvir Sajed , Daniel Johnson , Carin Li , Zinat Sayeeda , et al. Drugbank 5.0: a major update to the drugbank database for 2018 . Nucleic acids research , 46 ( D1 ): D1074 – D1082 , 2018 . OpenUrl CrossRef PubMed 41. ↵ Guangchuang Yu , Li-Gen Wang , Yanyan Han , and Qing-Yu He . clusterprofiler: an r package for comparing biological themes among gene clusters . Omics: a journal of integrative biology , 16 ( 5 ): 284 – 287 , 2012 . OpenUrl CrossRef PubMed 42. ↵ Shuyu Zheng , Jehad Aldahdooh , Tolou Shadbahr , Yinyin Wang , Dalal Aldahdooh , Jie Bao , Wenyu Wang , and Jing Tang . Drugcomb update: a more comprehensive drug sensitivity data repository and analysis portal . Nucleic acids research , 49 ( W1 ): W174 – W184 , 2021 . OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted August 14, 2024. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening Jingyu Yang , Meng Wang , Jürgen Dönitz , Björn Chapuy , Tim Beißbarth medRxiv 2024.06.20.24309094; doi: https://doi.org/10.1101/2024.06.20.24309094 Share This Article: Copy Citation Tools Advancing Personalized Cancer Therapy: Onko_DrugCombScreen - A Novel Shiny App for Precision Drug Combination Screening Jingyu Yang , Meng Wang , Jürgen Dönitz , Björn Chapuy , Tim Beißbarth medRxiv 2024.06.20.24309094; doi: https://doi.org/10.1101/2024.06.20.24309094 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Oncology Subject Areas All Articles Addiction Medicine (574) Allergy and Immunology (865) Anesthesia (304) Cardiovascular Medicine (4462) Dentistry and Oral Medicine (445) Dermatology (383) Emergency Medicine (611) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1517) Epidemiology (15251) Forensic Medicine (31) Gastroenterology (1132) Genetic and Genomic Medicine (6621) Geriatric Medicine (669) Health Economics (1002) Health Informatics (4564) Health Policy (1372) Health Systems and Quality Improvement (1617) Hematology (544) HIV/AIDS (1272) Infectious Diseases (except HIV/AIDS) (15938) Intensive Care and Critical Care Medicine (1107) Medical Education (624) Medical Ethics (147) Nephrology (670) Neurology (6643) Nursing (346) Nutrition (1001) Obstetrics and Gynecology (1149) Occupational and Environmental Health (957) Oncology (3350) Ophthalmology (981) Orthopedics (369) Otolaryngology (421) Pain Medicine (436) Palliative Medicine (130) Pathology (665) Pediatrics (1698) Pharmacology and Therapeutics (694) Primary Care Research (714) Psychiatry and Clinical Psychology (5465) Public and Global Health (9259) Radiology and Imaging (2212) Rehabilitation Medicine and Physical Therapy (1372) Respiratory Medicine (1198) Rheumatology (598) Sexual and Reproductive Health (716) Sports Medicine (533) Surgery (715) Toxicology (100) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a03d0bdd9b4d8df8',t:'MTc4MDEzNjg0Ng=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();