Radiogenomics and Machine Learning Predict Oncogenic Signaling Pathways in Glioblastoma

Radiogenomics and Machine Learning Predict Oncogenic Signaling Pathways in Glioblastoma | 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 Radiogenomics and Machine Learning Predict Oncogenic Signaling Pathways in Glioblastoma Abdul Basit Ahanger, Syed Wajid Aalam, Tariq Ahmad Masoodi, Asma Shah, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5131289/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Glioblastoma (GBM) is a highly aggressive brain tumor associated with a poor patient prognosis. Despite standard therapies, the survival rate remains low, highlighting the urgent need for novel treatment strategies. Advanced imaging techniques, particularly magnetic resonance imaging (MRI), play a crucial role in assessing GBM. Disruptions in various oncogenic signalling pathways, such as Receptor Tyrosine Kinase (RTK)-Ras-Extracellular signal-regulated kinase (ERK) signalling, Phosphoinositide 3- Kinases (PI3Ks), tumour protein p53 (TP53), and Neurogenic locus notch homolog protein (NOTCH), contribute to the development of different tumour types, each exhibiting distinct morphological and phenotypic features that can be observed at a microscopic level. However, identifying genetic abnormalities for targeted therapy often requires invasive procedures, prompting exploration into non-invasive approaches like radiogenomics. This study explores the utility of radiogenomics and machine learning (ML) in predicting these oncogenic signaling pathways in GBM patients. Data from MRI scans and signaling pathways were collected, radiomic features were extracted, and ML models were trained and evaluated using cross-validation techniques. Our results showed a positive association between most signalling pathways and the radiomic features derived from MRI scans. The best models achieved high AUC scores, namely 0.7 for RTK-RAS, 0.8 for PI3K, 0.75 for TP53, and 0.4 for NOTCH, and therefore demonstrated the potential of ML models in accurately predicting oncogenic signaling pathways from radiomic features, thereby informing personalized therapeutic approaches and improving patient outcomes. We present a novel approach for the non-invasive prediction of deregulation in oncogenic signaling pathways in glioblastoma (GBM) by integrating radiogenomic data with machine learning (ML) models. This research contributes to the advancement of precision medicine in GBM management, highlighting the importance of integrating radiomics with genomic data to better understand tumor behavior and treatment response. Glioblastoma Signalling Pathways Radiogenomics Machine Learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Glioblastoma (GBM), a formidable malignancy originating from neural tissue within the brain, presents a significant challenge in oncology, representing nearly half of all primary brain tumors. Tragically, the prognosis for GBM patients remains bleak, with an average survival period of 15 to 20 months and a mere 5% exhibiting potential for remission over 3 to 5 years [1-6]. GBM cells exhibit elevated aggressiveness, resistance to standard therapies, and rapid proliferation, predominantly localized in the supratentorial region and frontal lobe [1, 3, 7]. Classified as a grade 4 malignancy by the World Health Organization (WHO), GBM demands urgent attention due to its devastating impact, especially in older individuals, with a median age of diagnosis at 64 years [8-11]. Over recent years, the incidence of GBM in the USA has been on the rise, with an estimated 14,190 new cases reported in 2022, reflecting a 12.1% increase compared to previous years. Despite following the standard clinical protocol involving surgical resection, chemotherapy, and radiotherapy, GBM patients continue to face a meager 5-year survival rate of only 6.9%, highlighting the urgent need for novel treatment strategies [12, 13]. Advanced imaging techniques such as magnetic resonance imaging (MRI) play a crucial role in accurately assessing glioblastoma. While these imaging modalities have greatly improved diagnosis, the identification of genetic abnormalities for targeted therapy often necessitates invasive procedures, carrying significant risks [14-16]. A promising solution to these challenges lies in radiogenomics, a non-invasive approach that analyzes radiomic features extracted from various imaging modalities to detect and identify tumors. Radiogenomics combines genetics, radiomics, and artificial intelligence (AI), offering new possibilities for targeted therapy and precision medicine by enabling the identification of genetic profiles. By analyzing phenotypic data obtained from medical imaging, radiogenomics provides insights into tumor behavior and treatment response [17-20]. Recent studies have shown direct associations between tumors' phenotypic and genotypic characteristics, sparking interest in extracting genotypic information from medical imaging techniques [21-23]. The disruption of signaling pathways such as Ras-ERK (extracellular signal-regulated kinase) and PI3K (phosphatidylinositol 3-kinase) can lead to abnormal cell behavior and contribute to cancer progression [24-26]. Targeted inhibition of these pathways using precision drugs offers a personalized treatment strategy [27, 28]. Glioblastoma, a formidable opponent in cancer research, effectively utilizes various signaling pathways to fuel the continuous growth and survival of cancerous cells. This coordinated network plays a crucial role in sustaining the aggressive behavior of the tumor. One of the main pathways is PI3K, which plays a vital role in glioblastoma's pathophysiology. The PI3K pathway often becomes overactive due to the loss of the PTEN ( Phosphatase and Tensin Homolog) tumor suppressor gene, promoting tumor growth [29]. Additionally, NOTCH (Neurogenic Locus Notch Homolog Protein) signaling, which is essential in gliomagenesis, interacts with the TP53 pathway, affecting cell death [30-33]. Unfortunately, TP53 (Tumor protein 53), a vital tumor suppressor, is frequently disrupted in GBM [29]. Furthermore, The Receptor tyrosine kinase (RTK-RAS) pathway, commonly altered in glioblastoma, is linked to overactive RTK receptors, further fueling the complexity of the disease [34-36]. Moreover, WNT signaling, which is essential for CNS development, becomes dysregulated in GBM, especially in glioblastoma stem cells (GSCs), which contributes to tumor growth and makes it resistant to drugs [37-40]. Traditionally, the identification of these pathways relies upon gene expression analysis, a laborious and resource-intensive process [41]. However, the advent of radiomics holds immense promise in revolutionizing GBM management. By utilizing imaging modalities like MRI, radiomics facilitates non-invasive elucidation of underlying molecular processes, potentially offering personalized treatment strategies and substantially enhancing patient outcomes. In this context, our study explores the utility of radiogenomics and machine learning (ML) in predicting oncogenic signaling pathways in GBM. By extracting radiomic features from MRI scans and integrating them with genomic data, we aim to identify associations between imaging phenotypes and genetic profiles, mainly focusing on key signaling pathways implicated in gliomagenesis. Through advanced ML algorithms, we seek to develop predictive models that accurately identify these pathways, thereby informing personalized therapeutic approaches and improving patient outcomes. Methodology Data Collection of MRI Scans and Signalling Pathways Our experimental setup commenced by acquiring post-operative scans with multiple parameters from various modalities, comprising T1-weighted (T1w), contrast-enhanced T1-weighted (T1c), fluid-attenuated inversion recovery (FLAIR), and T2-weighted (T2w) images. These scans were sourced from the BRATS-19 dataset, made available by the Centre for Biomedical Image Computing & Analytics (CBCIA) at the University of Pennsylvania [42-44]. The dataset comprised MRI scans obtained from patients diagnosed with both GBM and Low-Grade Glioma (LGG), accompanied by publicly accessible genomic and other clinical data, which can be accessed through platforms such as The Cancer Genome Atlas (TCGA) [35] and Clinical Proteomic Tumour Analysis Consortium (CPTAC) [45]. CBCIA offers a file-name mapping that correlates the provided scans with patient identifiers in TCGA and CPTAC portals, facilitating access to genetic and clinical information from alternative sources. The signaling pathways of the relevant patients were gathered from cBioPortal. This platform offers interactive access to genomic profiles across different datasets and hosts datasets of signaling pathways associated with these profiles. Due to the absence of an API for accessing the pathway dataset, the pathway data was manually extracted using a web scraper. This extraction was performed from the GBM TCGA PanCancer Atlas (Study ID: gbm_tcga_pan_can_atlas_2018), GBM CPTAC (Study ID: gbm_cptac_2021), and Brain Lower Grade Glioma TCGA PanCancer (Study ID: gbm_tcga_pan_can_atlas_2018) datasets. In cases where certain pathways were unavailable in these datasets, they were obtained alternatively from the TCGA Firehose Legacy dataset (Study ID: gbm_tcga, lgg_tcga). While the training set encompassed MRI scans of LGG, it's noteworthy that GBM and LGG exhibit unique features and certain discrepancies in appearance. However, despite these disparities, the two types of gliomas also display overlapping imaging features and share specific genetic characteristics. Both tumors show increased signal intensity in T2-weighted scans, decreased intensity at T1-weighted imaging, and exhibit contrast enhancement when a contrasting agent is used [46]. The MRI Scans in both cases show infiltrative growth patterns by exhibiting abnormal signal intensity beyond the tumor mass to indicate the spread of the tumor [47]. They show similarities in genetic mutation [48, 49] and also have similarities in the aberrations of their signaling pathways [50, 51]. Radiomic Feature Extraction of MRI Scans A panel of radiomic features was extracted from the four MRI modalities using the segmentation mask provided for each subject. The extraction process was conducted utilizing PyRadiomics [52], which includes feature definitions that are compliant with the Imaging Biomarker Standardization Initiative (IBSI) [53]. The IBSI standardizes feature definitions and furnishes reference values for verifying radiomic software, enhancing reproducibility, and facilitating the clinical translation of radiomic research. The extracted feature sets underwent standardization to achieve a normal distribution and normalization within the range [0,1]. However, the abundance of extracted features, numbering over a thousand (>1000), could potentially lead to the "curse of dimensionality." [54]. Dimensionality reduction and selection of the most relevant features for the outcome model were performed through feature engineering on the generated feature panel. Data Balancing To address the data imbalance issue in our dataset, we employed the Synthetic Minority Oversampling Technique (SMOTE) [55]. SMOTE utilizes the K-Nearest Neighbors algorithm to identify neighboring feature vectors within the minority class in a given vector space. Subsequently, it generates synthetic data points along the line connecting these neighboring feature vectors of the minority class. This results in an expanded and less narrowly defined decision boundary between the classes, consequently enhancing the performance of classification models. The newly created instance stays within the original feature space, maintaining the dataset's characteristics and distribution. Furthermore, the generated data introduces extra decision boundaries for a more generalized model, unlike methods such as random oversampling, which merely replicates existing instances. Model Selection and Training Five supervised classification models based on ML were chosen and trained using the provided feature set to forecast the five signaling pathways. The selected algorithms comprise a Logistic Regression Classifier (LRC), Support Vector Machine (SVM), Random Forest Classifier (RFC), AdaBoost Classifier (ABC), and K-Nearest Neighbor Classifier (KNN). To enhance the prediction accuracy of our models, we employed Grid Search for hyperparameter tuning, aiming to determine the optimal hyperparameter configurations for each algorithm. The models underwent 5-fold cross-validation to ensure a more precise and unbiased performance assessment. Subsequently, the models with the most suitable hyperparameter settings were evaluated on separate, unseen test sets. Various evaluation metrics were used to identify each signaling pathway's top-performing algorithms. Results We examine the results of our experiments through cross-validation using four classification algorithms: RFC, SVC, ABC, and LRC. The objective was to detect five specific signaling pathways (WNT, PI3K, TP53, RTK-RAS, NOTCH) across three different datasets: over_sample, under_sample, and under_sample_pure. Our approach involved employing a 5-fold cross-validation technique and evaluating the algorithms' performance based on accuracy, precision, recall, and F1-score. MRI Scans and Segmentation Labels The multi-institutional pre-operative MRI scans for GBM and LGG patients of TCGA (n =167) and CPTAC (n =19) were made available in the BRATS-19 dataset released by CBCIA. In addition to containing T1, T2, Post Contrast T1, and FLAIR 3D MRI volumes, the dataset comprises segmentation labels experienced neuro-radiologists have verified. These segmentation labels include annotations for the Enhancing Tumor part (ET), Necrotic and non-enhancing Tumor part (NET), and Peritumoral Edema (ED). The scans have undergone various preprocessing steps, including skull-stripping, co-registration, and interpolation, to achieve a resolution of 1 mm³. Collection of Signaling Pathways and Mapping with MRI Scans Nine oncogenic signaling pathways listed in Table 1 of the feature set were extracted from cBioPortal. However, two pathways, NRF2 and TGF-β, exhibited no alterations and were consequently excluded from further analysis. As depicted in Figure 2 , the distribution of pathway alterations is imbalanced, with either an excess or a scarcity of alterations observed. In situations with a significant disparity between the majority and minority classes, ML algorithms tend to bias their classification outcomes towards the majority class, leading to skewed results [56]. While achieving high accuracy, the algorithm may not perform optimally regarding other performance metrics like sensitivity (recall) or F1-Score. Despite its strong performance, SMOTE may not effectively handle severe imbalances in the data. To tackle this challenge, the top four oncogenic signaling pathways with the least imbalance (<30%) were chosen for model training. These pathways include PI3K, TP53, RTK-RAS, WNT, and NOTCH signaling pathways, which have been demonstrated to impact GBM significantly [29-34, 36-41]. The RTK-RAS, PI3K, NOTCH, TP53, and WNT signaling pathways are interconnected, with their interactions and cross-regulation playing a substantial role in the development and advancement of GBM [29-34, 36-41]. Recognizing these signaling pathways and understanding the crosstalk between them is essential for administering and advancing targeted therapies. The relationships among signaling pathways across the subjects are depicted in an upset plot in Figure 3 . Derivation of Radiomic Feature Panel A comprehensive radiomic feature set of 1284 features, derived from 101 standard features following IBSI standards, was obtained from the four imaging sequences (T1, T2, T1c, FLAIR) along with their corresponding segmentation masks. The extracted features encompassed first-order, volumetric, and intensity-based textural features categorized as First Order, Shape, GLCM, GLDM, GLRLM, GLSZM, and NGTDM, with all 107 features listed in Table 2 . Dimensionality Reduction and Feature Selection To diminish the dimensionality of the feature set, features exhibiting significant correlation with the target pathway (>0.1) and minimal correlation with each other (<0.9) were chosen from the radiomic feature panel. Subsequently, feature importance was assessed for each training set using the random forest algorithm in sklearn with default parameter settings to identify the top 10 most important features. The top 10 features selected for each pathway are presented in Table 3 . Creation of Datasets After preprocessing, feature engineering, and data splitting, the final data cohort consists of 167 subjects from TCGA (GBM =98, LGG =69) and 19 from CPTAC. It would’ve been suitable to set the 19 subjects from CTPAC aside for data validation; however, due to heavy imbalance in the signaling pathways of the CPTAC dataset, the experimentation was carried out on three separate datasets created by splitting the cohort into training and validation sets ( Figure 4 ). i) In the over_split dataset, the validation set was created by combining samples from the CPTAC dataset with some samples from TCGA. The selection of TCGA samples depended on the availability of samples for the minority class in TCGA. There were specific criteria for constructing this validation set. If the minority class in the CPTAC dataset represented the majority class in TCGA-GBM, then all samples of the minority class from the training set were incorporated into the validation set. Similarly, if the minority class was consistent between TCGA-GBM and CPTAC, samples from TCGA-GBM were chosen for inclusion in the validation set. The number of cases transferred to the validation set depended on the ratio of the majority class to the minority class in TCGA-GBM. If this ratio was less than 0.11, three cases were moved to the validation set; otherwise, five cases were moved to ensure balanced representation. This was done to balance the class distribution. ii) The under_split dataset features a validation set comprising solely CPTAC samples, balanced using under-sampling techniques. This involved restricting the number of samples in the majority class to twice that of the minority class. The remaining samples from the majority class were then utilized to augment the size of the training set, thereby enhancing the training process. iii) The under_split_pure dataset encompassed a validation set comprising exclusively CPTAC samples, which were balanced via under-sampling. This entailed restricting the number of samples in the majority class to twice that of the minority class, with any surplus samples being discarded. Evaluation of ML Models using K-Fold Cross Validation Five ML algorithms were trained to detect five fundamental oncogenic signaling pathways by training them on diverse radiomic features extracted from segmentation labels of GBM MRI scans. The models underwent training using five-fold cross-validation to assess the generalizability of the employed approach. The mean results of the 5-fold cross-validation of each model across all three datasets in terms of accuracy are tabulated in Table 4 and further visualized for clarity in Figure 5 . In addition to accuracy, the ROC_AUC score was selected to provide a more comprehensive assessment of the models' performance. The mean results here showcase the optimal parameters achieved through thorough hyperparameter tuning using Grid Search. The precise hyperparameter configuration responsible for these results is outlined in Table 5 . Model Validation on Unseen Data The model that performed the best on each signaling pathway was further validated on the test sets set aside to evaluate the generalizability of the models. The following metrics were chosen to give an idea of the general prediction power, showcase the performance on imbalanced data, and observe the trade-off between the correct prediction rate and misclassifications. Accuracy may not be the most suitable metric in classification problems when dealing with moderate to severe data imbalance. In such scenarios, precision, recall (specificity), and F1-score offer a more precise assessment of the models' performance. Comparative visualizations of the algorithms based on accuracy, precision, recall, and F1-score are presented in Figure 6. Discussion and Conclusion Understanding the interconnectedness of signaling pathways like RTK-RAS, PI3K, NOTCH, and TP53 is vital for comprehending how they influence GBM's development and advancement. Detecting these pathways and their mutual communication is pivotal for administering existing treatments and advancing targeted therapies. Interesting associations can also be seen in our data and shown in the upset plot in Figure 3 . PI3K, RTK-RAS, or TP53 alterations occur independently in no more than three cases each, whereas the NOTCH pathway was observed separately in nine cases and the WNT pathway in 16 cases. Conversely, PI3K, RTK-RAS, and TP53 are often co-present, with various combinations observed. For instance, unique combinations of RTK-RAS and TP53 were found in 31 instances, TP53 with only PI3K in 14 cases, and all three together in 30 cases. These three pathways are well-known for gliomagensis [35, 57, 58]. In addition to the simultaneous presence of established pathways, there are instances where pathways such as NOTCH, which is recognized for inducing apoptosis in the P53 gene, are found alongside the TP53 pathway in 35 cases. The interaction between these pathways is extensively studied and documented [33, 59]. Notably, all four pathways are present in just seven out of the total 167 cases. The research highlights the significance of carefully selecting algorithms, ensuring data quality, and conducting thorough feature engineering in radio-genomic studies aimed at pathway detection. Each pathway demonstrates distinct interactions and presents specific challenges in this regard. Various ML models demonstrate varying evaluation scores, suggesting that particular pathways may present more complex classification tasks than others. For example, the PI3K pathways seemed more difficult for models to classify accurately, as evidenced by lower precision, recall, and F1-score values. Variations in class distribution also influence model performance, as different representations of classes introduce biases in learning. Specific pathways, like NOTCH and PI3K, seemed to pose more significant challenges for models than others, as reflected in lower precision, recall, and F1-score values. This suggests that the classification boundaries within these pathways are more complex. The varied performance observed among different ML algorithms corresponds with the distinct challenges presented by each pathway and the size of the dataset. Ensemble methods, notably Random Forest, demonstrated consistent performance across different situations, indicating their potential as dependable baseline models. The TP53 pathway, renowned for its function as a tumor suppressor, yielded exciting findings. When applied to the over_split dataset, the ABC algorithm showed remarkable accuracy, precision, and F1-score, indicating its efficacy in detecting this pathway. Yet, the algorithm's efficacy dramatically declined on the under_split and under_split_pure datasets, with most algorithms showing exceptionally low precision and recall, likely due to the minimal test set comprising only three sample points. Conversely, cross-validation accuracy on these two datasets remained consistently above 0.70, except for LRC. The RTK-RAS pathway, characterized by its intricate network of interactions, displayed diverse performance across different datasets. In the over_split dataset, the SVC exhibited a balanced performance. However, except KNN, all algorithms failed to identify any true negatives, classifying all nine samples as positives, leading to 100% recall but zero precision. This discrepancy contrasts with the cross-validation outcomes on the training set, suggesting that none of the models have overfit and have not achieved successful generalization. The PI3K pathway, which plays a crucial role in cell growth and survival, exhibited relatively consistent performance trends on under_split_pure and over_split datasets. In the case of over_split, the RFC achieved consistent results across all metrics, reflecting its robustness. However, on the under_split dataset, all algorithms faced challenges detecting any true negatives primarily due to possible class imbalance issues. Interestingly, the RFC algorithm excelled on the under_split_pure dataset, with all 9 cases correctly detected. The NOTCH signaling pathway is known for its significance in gliomagenesis. Across all datasets, we observed consistently higher performance among all the algorithms over all three datasets in the prediction of NOTCH Pathway. The LRC displayed exceptional precision and recall on the over_split and under_split datasets. In contrast, all algorithms failed to perform significantly on the under_split_pure dataset. The WNT signaling pathway, critical for cell differentiation in the central nervous system, presented diverse performance across algorithms. We had the highest class imbalance and the least number of alterations among the pathways, which performed relatively poorly on all the datasets. RFC showed variable performance across datasets, while the LRC achieved high precision and recall on the under_split dataset. The SVC demonstrated strong performance on the under_split_pure dataset, indicating its capability to handle class imbalance effectively. This research revealed insights into utilizing machine learning models with radiomic data to forecast specific oncogenic signaling pathways. The findings underscore the impact of dataset size, class distribution, and feature complexity on model effectiveness. By considering these elements, we can enhance our prediction algorithms, fostering a deeper understanding of employing AI in radiomics to elucidate the interactions among different signaling pathways and their influence on tumor phenotypic traits. The prediction of oncogenic signaling pathways from radiomic features holds promise for advancing genomic diagnosis faster and more cost-effectively. Invasive diagnostic procedures for brain tumors, such as brain biopsies, entail additional risks, making the timely and accurate genetic profiling of specimens crucial for targeted therapeutic interventions in Glioblastoma cases. Our study deployed four machine-learning models to forecast four oncogenic signaling pathways using MRI scans from the TCGA-GBM dataset. Our findings revealed a positive correlation between the radiomic features extracted from MRI scans and oncogenic signaling pathways in GBM. With adequate data, manual feature extraction could be bypassed, leading to the development of a more generalized multi-label deep learning model capable of predicting additional signaling pathways. We intend to expand this research by developing a multi-label deep learning model that can predict a broader spectrum of signaling pathways. Declarations Ethical Approval and Consent to participate Not Applicable. Consent for publication Not Applicable. Availability of supporting data Not Applicable. Competing interests The authors declare that they have no competing interests. Conflict of Interest Statement Author MK was employed by DigiBiomics Inc. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Acknowledgments Funding This study was funded, in part, by a Research Grant (Grant number: ID No. 2022-16465) from the Indian Council of Medical Research (ICMR) Govt. of India, New Delhi to Muzafar A. Macha. Promotion of University Research and Scientific Excellence (PURSE) (SR/PURSE/2022/121) grant from the Department of Science and Technology, Govt. of India, New Delhi to the Islamic University of Science and Technology (IUST), Awantipora. Author contributions ABA, SWA, MRB, and MAM wrote the manuscript and generated figures. TAM, AA, MAM, and MRB contributed to the concept & design and critically edited the manuscript. ABA, SWA and TAM performed experiments. ABA, SWA, TAM, AS, MAK, AAB, AS, MRB, and MAM critically revised and edited the scientific content. All authors read and approved the final manuscript. References Grochans S, Cybulska AM, Simińska D, Korbecki J, Kojder K, Chlubek D, Baranowska-Bosiacka I: Epidemiology of glioblastoma multiforme–literature review . Cancers 2022, 14 (10):2412. Lu I-N, Dobersalske C, Rauschenbach L, Teuber-Hanselmann S, Steinbach A, Ullrich V, Prasad S, Blau T, Kebir S, Siveke JT: Tumor-associated hematopoietic stem and progenitor cells positively linked to glioblastoma progression . Nature communications 2021, 12 (1):3895. Kim HJ, Park JW, Lee JH: Genetic architectures and cell-of-origin in glioblastoma . Frontiers in oncology 2021, 10 :615400. Melhem JM, Detsky J, Lim-Fat MJ, Perry JR: Updates in IDH-wildtype glioblastoma . Neurotherapeutics 2022, 19 (6):1705-1723. Tykocki T, Eltayeb M: Ten-year survival in glioblastoma. A systematic review . Journal of Clinical Neuroscience 2018, 54 :7-13. Patel NP, Lyon KA, Huang JH: The effect of race on the prognosis of the glioblastoma patient: a brief review . Neurological research 2019, 41 (11):967-971. Bai J, Varghese J, Jain R: Adult glioma WHO classification update, genomics, and imaging: what the radiologists need to know . Topics in Magnetic Resonance Imaging 2020, 29 (2):71-82. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, Hawkins C, Ng H, Pfister SM, Reifenberger G: The 2021 WHO classification of tumors of the central nervous system: a summary . Neuro-oncology 2021, 23 (8):1231-1251. McNeill KA: Epidemiology of brain tumors . Neurologic clinics 2016, 34 (4):981-998. Ostrom QT, Gittleman H, Farah P, Ondracek A, Chen Y, Wolinsky Y, Stroup NE, Kruchko C, Barnholtz-Sloan JS: CBTRUS statistical report: Primary brain and central nervous system tumors diagnosed in the United States in 2006-2010 . Neuro-oncology 2013, 15 (suppl_2):ii1-ii56. Chakrabarti I, Cockburn M, Cozen W, Wang YP, Preston‐Martin S: A population ‐based description of glioblastoma multiforme in Los Angeles County, 1974–1999 . Cancer: Interdisciplinary International Journal of the American Cancer Society 2005, 104 (12):2798-2806. Ostrom QT, Gittleman H, Truitt G, Boscia A, Kruchko C, Barnholtz-Sloan JS: CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2011–2015 . Neuro-oncology 2018, 20 (suppl_4):iv1-iv86. Ostrom QT, Price M, Neff C, Cioffi G, Waite KA, Kruchko C, Barnholtz-Sloan JS: CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2015-2019 . Neuro Oncol 2022, 24 (Suppl 5):v1-v95. Bi WL, Hosny A, Schabath MB, Giger ML, Birkbak NJ, Mehrtash A, Allison T, Arnaout O, Abbosh C, Dunn IF: Artificial intelligence in cancer imaging: clinical challenges and applications . CA: a cancer journal for clinicians 2019, 69 (2):127-157. Hussain T, Nguyen QT: Molecular imaging for cancer diagnosis and surgery . Advanced drug delivery reviews 2014, 66 :90-100. Chen C-C, Hsu P-W, Wu T-WE, Lee S-T, Chang C-N, Wei K-c, Chuang C-C, Wu C-T, Lui T-N, Hsu Y-H: Stereotactic brain biopsy: Single center retrospective analysis of complications . Clinical neurology and neurosurgery 2009, 111 (10):835-839. Shui L, Ren H, Yang X, Li J, Chen Z, Yi C, Zhu H, Shui P: The era of radiogenomics in precision medicine: an emerging approach to support diagnosis, treatment decisions, and prognostication in oncology . Frontiers in Oncology 2021, 10 :570465. Bodalal Z, Trebeschi S, Nguyen-Kim TDL, Schats W, Beets-Tan R: Radiogenomics: bridging imaging and genomics . Abdominal radiology 2019, 44 (6):1960-1984. Sundaram M, McGuire MH, Herbold DR: Magnetic resonance imaging of soft tissue masses: an evaluation of fifty-three histologically proven tumors . Magnetic resonance imaging 1988, 6 (3):237-248. Upadhyay N, Waldman A: Conventional MRI evaluation of gliomas . The British journal of radiology 2011, 84 (special_issue_2):S107-S111. Zinn PO, Singh SK, Kotrotsou A, Hassan I, Thomas G, Luedi MM, Elakkad A, Elshafeey N, Idris T, Mosley J: A coclinical radiogenomic validation study: conserved magnetic resonance radiomic appearance of periostin-expressing glioblastoma in patients and xenograft models . Clinical Cancer Research 2018, 24 (24):6288-6299. Pei L, Vidyaratne L, Rahman MM, Iftekharuddin KM: Context aware deep learning for brain tumor segmentation, subtype classification, and survival prediction using radiology images . Scientific Reports 2020, 10 (1):19726. Li K, Xiao J, Yang J, Li M, Xiong X, Nian Y, Qiao L, Wang H, Eresen A, Zhang Z: Association of radiomic imaging features and gene expression profile as prognostic factors in pancreatic ductal adenocarcinoma . American journal of translational research 2019, 11 (7):4491. Sever R, Brugge JS: Signal transduction in cancer . Cold Spring Harbor perspectives in medicine 2015, 5 (4):a006098. Villanueva A, García C, Paules AB, Vicente M, Megías M, Reyes G, de Villalonga P, Agell N, Lluís F, Bachs O: Disruption of the antiproliferative TGF-β signaling pathways in human pancreatic cancer cells . Oncogene 1998, 17 (15):1969-1978. Mendoza MC, Er EE, Blenis J: The Ras-ERK and PI3K-mTOR pathways: cross-talk and compensation . Trends in biochemical sciences 2011, 36 (6):320-328. Samimi H, Fallah P, Sohi AN, Tavakoli R, Naderi M, Soleimani M, Larijani B, Haghpanah V: Precision medicine approach to anaplastic thyroid cancer: advances in targeted drug therapy based on specific signaling pathways . Acta Medica Iranica 2017:200-208. Zhu P, Aliabadi HM, Uludağ H, Han J: Identification of potential drug targets in cancer signaling pathways using stochastic logical models . Scientific reports 2016, 6 (1):23078. Mischel PS, Nelson SF, Cloughesy TF: Molecular analysis of glioblastoma: pathway profiling and its implications for patient therapy . Cancer biology & therapy 2003, 2 (3):242-247. Pearson JR, Regad T: Targeting cellular pathways in glioblastoma multiforme . Signal transduction and targeted therapy 2017, 2 (1):1-11. Bazzoni R, Bentivegna A: Role of notch signaling pathway in glioblastoma pathogenesis . Cancers 2019, 11 (3):292. Gaiano N, Fishell G: The role of notch in promoting glial and neural stem cell fates . Annual review of neuroscience 2002, 25 (1):471-490. Yang X, Klein R, Tian X, Cheng H-T, Kopan R, Shen J: Notch activation induces apoptosis in neural progenitor cells through a p53-dependent pathway . Developmental biology 2004, 269 (1):81-94. Sanchez-Vega F, Mina M, Armenia J, Chatila WK, Luna A, La KC, Dimitriadoy S, Liu DL, Kantheti HS, Saghafinia S: Oncogenic signaling pathways in the cancer genome atlas . Cell 2018, 173 (2):321-337. e310. CGARNTssDUMSMRFABD, 5 EUVMEGBDJMMGOJJ, 8 HFHMTLN, 11 MACCAKAYWBO, 13 UoCSFVSBMPM: Comprehensive genomic characterization defines human glioblastoma genes and core pathways . Nature 2008, 455 (7216):1061-1068. Fonseca COd, Linden R, Futuro D, Gattass CR, Quirico-Santos T: Ras pathway activation in gliomas: a strategic target for intranasal administration of perillyl alcohol . Archivum immunologiae et therapiae experimentalis 2008, 56 :267-276. Lee Y, Lee J-K, Ahn SH, Lee J, Nam D-H: WNT signaling in glioblastoma and therapeutic opportunities . Laboratory investigation 2016, 96 (2):137-150. Komiya Y, Habas R: Wnt signal transduction pathways . Organogenesis 2008, 4 (2):68-75. Guan R, Zhang X, Guo M: Glioblastoma stem cells and Wnt signaling pathway: Molecular mechanisms and therapeutic targets . Chinese Neurosurgical Journal 2020, 6 :1-6. Zuccarini M, Giuliani P, Ziberi S, Carluccio M, Di Iorio P, Caciagli F, Ciccarelli R: The role of Wnt signal in glioblastoma development and progression: a possible new pharmacological target for the therapy of this tumor . Genes 2018, 9 (2):105. Kabir MH, Patrick R, Ho JW, O’Connor MD: Identification of active signaling pathways by integrating gene expression and protein interaction data . BMC systems biology 2018, 12 :77-87. Menze BH, Jakab A, Bauer S, Kalpathy-Cramer J, Farahani K, Kirby J, Burren Y, Porz N, Slotboom J, Wiest R: The multimodal brain tumor image segmentation benchmark (BRATS) . IEEE transactions on medical imaging 2014, 34 (10):1993-2024. Bakas S, Akbari H, Sotiras A, Bilello M, Rozycki M, Kirby JS, Freymann JB, Farahani K, Davatzikos C: Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features . Scientific data 2017, 4 (1):1-13. Bakas S, Reyes M, Jakab A, Bauer S, Rempfler M, Crimi A, Shinohara RT, Berger C, Ha SM, Rozycki M: Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge . arXiv preprint arXiv:181102629 2018. Whiteaker JR, Halusa GN, Hoofnagle AN, Sharma V, MacLean B, Yan P, Wrobel JA, Kennedy J, Mani D, Zimmerman LJ: CPTAC Assay Portal: a repository of targeted proteomic assays . Nature methods 2014, 11 (7):703-704. Haydar N, Alyousef K, Alanan U, Issa R, Baddour F, Al-Shehabi Z, Al-Janabi MH: Role of Magnetic Resonance Imaging (MRI) in grading gliomas comparable with pathology: A cross-sectional study from Syria . Annals of Medicine and Surgery 2022, 82 :104679. Guillevin R, Herpe G, Verdier M, Guillevin C: Low-grade gliomas: The challenges of imaging . Diagnostic and Interventional imaging 2014, 95 (10):957-963. Aoki K, Nakamura H, Suzuki H, Matsuo K, Kataoka K, Shimamura T, Motomura K, Ohka F, Shiina S, Yamamoto T: Prognostic relevance of genetic alterations in diffuse lower-grade gliomas . Neuro-oncology 2018, 20 (1):66-77. Paugh BS, Qu C, Jones C, Liu Z, Adamowicz-Brice M, Zhang J, Bax DA, Coyle B, Barrow J, Hargrave D: Integrated molecular genetic profiling of pediatric high-grade gliomas reveals key differences with the adult disease . Journal of clinical oncology 2010, 28 (18):3061. Nasrolahi A, Azizidoost S, Radoszkiewicz K, Najafi S, Ghaedrahmati F, Anbiyaee O, Khoshnam SE, Farzaneh M, Uddin S: Signaling pathways governing glioma cancer stem cells behavior . Cellular Signalling 2023, 101 :110493. Nakada M, Kita D, Watanabe T, Hayashi Y, Teng L, Pyko IV, Hamada J-I: Aberrant signaling pathways in glioma . Cancers 2011, 3 (3):3242-3278. Van Griethuysen JJ, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, Beets-Tan RG, Fillion-Robin J-C, Pieper S, Aerts HJ: Computational radiomics system to decode the radiographic phenotype . Cancer research 2017, 77 (21):e104-e107. Zwanenburg A, Leger S, Vallières M, Löck S: Image biomarker standardisation initiative . arXiv preprint arXiv:161207003 2016. Limkin EJ, Sun R, Dercle L, Zacharaki EI, Robert C, Reuzé S, Schernberg A, Paragios N, Deutsch E, Ferté C: Promises and challenges for the implementation of computational medical imaging (radiomics) in oncology . Annals of Oncology 2017, 28 (6):1191-1206. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP: SMOTE: synthetic minority over-sampling technique . Journal of artificial intelligence research 2002, 16 :321-357. Krawczyk B: Learning from imbalanced data: open challenges and future directions . Progress in Artificial Intelligence 2016, 5 (4):221-232. Venkatesan S, Lamfers ML, Dirven CM, Leenstra S: Genetic biomarkers of drug response for small-molecule therapeutics targeting the RTK/Ras/PI3K, p53 or Rb pathway in glioblastoma . CNS oncology 2016, 5 (2):77-90. Han S, Wang P-F, Cai H-Q, Wan J-H, Li S-W, Lin Z-H, Yu C-J, Yan C-X: Alterations in the RTK/Ras/PI3K/AKT pathway serve as potential biomarkers for immunotherapy outcome of diffuse gliomas . Aging (Albany NY) 2021, 13 (11):15444. Dotto GP: Crosstalk of Notch with p53 and p63 in cancer growth control . Nature reviews Cancer 2009, 9 (8):587-595. Tables Table 1 . Signaling Pathway Alterations S. No Pathway Alteration 1 TP53 83.67% 2 NRF2 0.00% 3 NOTCH 19.39% 4 MYC 6.12% 5 WNT 14.29% 6 TGF-β 0.00% 7 RTK-RAS 85.71% 8 PI3K 43.88% 9 HIPPO 11.22% Table 2 . Feature List after Feature Selection NOTCH PI3K RTK-RAS TP53 TEXTURE-GLRLM-ET-T1Gd-SRE TEXTURE-GLRLM-ET-T1Gd-GLN TEXTURE-GLCM-ED-T1Gd-Energy HISTO-ED-T2-Bin7 TEXTURE-GLCM-NET-T2-Correlation SPATIAL-Parietal TEXTURE-GLOBAL-ED-T1Gd-Kurtosis INTENSITY-Mean-NET-FLAIR TEXTURE-GLSZM-ET-T2-LGZE VOLUME-NET-OVER-ED TEXTURE-GLOBAL-NET-T1-Variance HISTO-NET-FLAIR-Bin1 TEXTURE-GLCM-NET-T1-Contrast VOLUME-BRAIN HISTO-ET-T2-Bin2 TEXTURE-GLCM-ED-T1Gd-Contrast TEXTURE-GLRLM-ET-T1-LRHGE TEXTURE-GLSZM-NET-T1Gd-SZE TEXTURE-NGTDM-ED-T1Gd-Busyness TEXTURE-GLCM-NET-T1Gd-Correlation HISTO-ET-T1-Bin10 HISTO-ET-FLAIR-Bin1 TEXTURE-GLOBAL-ET-T1-Skewness INTENSITY-Mean-ED-FLAIR HISTO-NET-T1Gd-Bin5 TEXTURE-GLRLM-NET-FLAIR-HGRE TEXTURE-GLRLM-ET-FLAIR-SRE TEXTURE-GLOBAL-ET-T1Gd-Skewness HISTO-NET-FLAIR-Bin9 TEXTURE-GLCM-NET-FLAIR-SumAverage TEXTURE-GLSZM-ET-FLAIR-SZE HISTO-NET-T1-Bin7 TEXTURE-GLRLM-NET-T2-RLV TEXTURE-GLRLM-ET-T1Gd-SRE TEXTURE-GLSZM-ET-T2-HGZE TEXTURE-NGTDM-ED-T1Gd-Busyness HISTO-ET-T2-Bin1 HISTO-ET-FLAIR-Bin9 TEXTURE-GLCM-ET-T1-Homogeneity HISTO-ET-T2-Bin4 Table 3 . Hyper-parameter state Algorithm Hyperparameter Pathways NOTCH PI3K RTK-RAS TP53 AdaBoost algorithm SAMME.R SAMME.R SAMME SAMME learning_rate 0.1 0.001 0.01 0.01 n_estimators 1600 2000 1800 1800 Logistic Regression C 100 100 100 100 penalty l1 l1 l2 l2 solver liblinear liblinear liblinear liblinear Random Forest bootstrap TRUE TRUE TRUE TRUE criterion gini gini gini gini max_features 0.2 0.2 0.2 0.2 n_estimators 1600 1600 1600 2000 Support Vector Machine C 1 1 1 100 0.001 1 1 0.01 kernel rbf linear linear rbf Cite Share Download PDF Status: Posted Version 1 posted 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-5131289","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":377817770,"identity":"c563b1ea-e8e6-4bf6-8b2d-c471408bad13","order_by":0,"name":"Abdul Basit Ahanger","email":"","orcid":"","institution":"Islamic University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Abdul","middleName":"Basit","lastName":"Ahanger","suffix":""},{"id":377817771,"identity":"0b081020-2c5d-4f0a-bd91-e4bc12aa3fe0","order_by":1,"name":"Syed Wajid Aalam","email":"","orcid":"","institution":"Islamic University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Syed","middleName":"Wajid","lastName":"Aalam","suffix":""},{"id":377817772,"identity":"7cb754af-95a9-4946-8b48-c98fcac2e3c6","order_by":2,"name":"Tariq Ahmad Masoodi","email":"","orcid":"","institution":"Sidra Medicine","correspondingAuthor":false,"prefix":"","firstName":"Tariq","middleName":"Ahmad","lastName":"Masoodi","suffix":""},{"id":377817773,"identity":"f35f7804-5943-48f8-96e4-3379074f8f87","order_by":3,"name":"Asma Shah","email":"","orcid":"","institution":"Islamic University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Asma","middleName":"","lastName":"Shah","suffix":""},{"id":377817774,"identity":"61814a68-c6ec-4ad5-aeac-5b00fbdf6f55","order_by":4,"name":"Meraj Alam Khan","email":"","orcid":"","institution":"digibiomics","correspondingAuthor":false,"prefix":"","firstName":"Meraj","middleName":"Alam","lastName":"Khan","suffix":""},{"id":377817775,"identity":"d1a4f43a-d060-465c-993d-6bb114324e1b","order_by":5,"name":"Ajaz A. Bhat","email":"","orcid":"","institution":"Sidra Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ajaz","middleName":"A.","lastName":"Bhat","suffix":""},{"id":377817776,"identity":"14e17d8b-8754-4180-8312-7fbce16a3f9d","order_by":6,"name":"Assif Assad","email":"","orcid":"","institution":"Islamic University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Assif","middleName":"","lastName":"Assad","suffix":""},{"id":377817777,"identity":"69b39e0c-6ea0-4615-aad1-f0ba91a96b8c","order_by":7,"name":"MUZAFAR A. MACHA","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8UlEQVRIiWNgGAWjYLCCChDBzNhwgIHBBshibDxAUMsZhJY0kJYGIrVAwGEwiVeL7oz0hx8O1Nxj4G9nbjx04895u7Xth4G21NhE49JidiMhWeLAsWIGicOMDYdz224nbzuTCNRyLC23AbeWA9If2BKATgJpabidbHYAqAXMxqklsfnHgX8JDPIgLTl/ziWbnX9ISEsym8TBtgQGA7AWtgN2QEMIaDnzjM3iYF8CjyHEL8kJZjeAtiTg88vx9Mc3DnxLkJM7f/zx55w/dvZm59MfPvhQY4NTCwzwwBiJYJUJBJSjAHtSFI+CUTAKRsHIAAD6I2vIuZHW2QAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-4468-4435","institution":"Islamic University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"MUZAFAR","middleName":"A.","lastName":"MACHA","suffix":""},{"id":377817778,"identity":"da168f37-9df7-4d47-a118-362b91fa8a41","order_by":8,"name":"Muzafar Rasool Bhat","email":"","orcid":"","institution":"Islamic University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Muzafar","middleName":"Rasool","lastName":"Bhat","suffix":""}],"badges":[],"createdAt":"2024-09-22 07:26:56","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5131289/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5131289/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":69362385,"identity":"338fe6cc-3c1e-42fc-a8e8-73c07b788dc7","added_by":"auto","created_at":"2024-11-19 14:31:42","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":101158,"visible":true,"origin":"","legend":"\u003cp\u003eGlioblastoma in various imaging modalities. The tumor part in T1w appears darker, while CSF is highlighted when contrast is added in T1c. The Tumour part appears brighter in T2 and FLAIR sequences.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/d6ace3fc5f90e41e0bd13901.png"},{"id":69363653,"identity":"134580d3-5573-4def-9f28-f488e3dbcf95","added_by":"auto","created_at":"2024-11-19 14:39:42","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":229663,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of alterations in 9 signaling pathways of all subjects in the cohort. The figure shows whether the pathways have alterations of more than half the size of the cohort. Only TP53 and RTK-RAS have more positive alterations than negative alterations. The pie charts show the percentage of positive and negative alterations of the five selected pathways.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/7e60838ba9cbfada02ef155c.png"},{"id":69363654,"identity":"b8459fe8-e9ef-466c-aa59-14cf2b45a9a1","added_by":"auto","created_at":"2024-11-19 14:39:42","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":196312,"visible":true,"origin":"","legend":"\u003cp\u003eUpset plot showing similarities in Signalling Pathways. Note that there are no intersections among WNT/NOTCH/PI3K, WNT/NOTCH/RTK-RAS, and WNT/NOTCH/PI3K/TP53.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/e2a9653e89bb371fa0c9c84c.png"},{"id":69362387,"identity":"d3431448-2ad7-4f2a-a604-90dd57965dc7","added_by":"auto","created_at":"2024-11-19 14:31:42","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":338818,"visible":true,"origin":"","legend":"\u003cp\u003eThe training and validation data distribution of the three datasets over_split, under_split and under_split_pure.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/33629782900940dcde6616aa.png"},{"id":69362390,"identity":"2a24b289-f8e3-4aab-a768-9083edafead7","added_by":"auto","created_at":"2024-11-19 14:31:42","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":542277,"visible":true,"origin":"","legend":"\u003cp\u003eMean 5-fold Cross-Validation accuracies on the training set for the five signaling pathways with standard deviation represented as error lines over each bar.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/6bd171dcf62aba8f02072067.png"},{"id":69362391,"identity":"5d8fe8e1-478e-42cb-8e76-a0165486289f","added_by":"auto","created_at":"2024-11-19 14:31:42","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2658903,"visible":true,"origin":"","legend":"\u003cp\u003eValidation result in terms of accuracy, precision, recall and F1-Score for (a) NOTCH, (b) PI3K, (c) RTK-RAS, (d) TP53 (e) WNT Signalling Pathways on AdaBoost Classifier (ABC), K-Nearest Neighbour (KNN), Logistic Regression Classifier (LRC), Random Forest Classifier (RFC), Support Vector Classifier (SVC)\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/c5672a258664662d630d1128.png"},{"id":69362389,"identity":"da1f5564-0fe8-4fc6-b295-3b33ac40c177","added_by":"auto","created_at":"2024-11-19 14:31:42","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":231901,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/fb7d464f0b099247e25c5536.png"},{"id":69364910,"identity":"f9077f2b-853b-4639-bdb9-6d13f636cfec","added_by":"auto","created_at":"2024-11-19 14:55:47","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7030607,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5131289/v1/33f63480-2896-4667-b990-eabebf2a224c.pdf"}],"financialInterests":"","formattedTitle":"Radiogenomics and Machine Learning Predict Oncogenic Signaling Pathways in Glioblastoma","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGlioblastoma (GBM), a formidable malignancy originating from neural tissue within the brain, presents a significant challenge in oncology, representing nearly half of all primary brain tumors. Tragically, the prognosis for GBM patients remains bleak, with an average survival period of 15 to 20 months and a mere 5% exhibiting potential for remission over 3 to 5 years\u0026nbsp;[1-6].\u0026nbsp;GBM cells exhibit elevated aggressiveness, resistance to standard therapies, and rapid proliferation, predominantly localized in the supratentorial region and frontal lobe\u0026nbsp;[1, 3, 7]. Classified as a grade 4 malignancy by the World Health Organization (WHO), GBM demands urgent attention due to its devastating impact, especially in older individuals, with a median age of diagnosis at 64 years\u0026nbsp;[8-11]. Over recent years, the incidence of GBM in the USA has been on the rise, with an estimated 14,190 new cases reported in 2022, reflecting a 12.1% increase compared to previous years. Despite following the standard clinical protocol involving surgical resection, chemotherapy, and radiotherapy, GBM patients continue to face a meager 5-year survival rate of only 6.9%, highlighting the urgent need for novel treatment strategies\u0026nbsp;[12, 13].\u003c/p\u003e\n\u003cp\u003eAdvanced imaging techniques such as magnetic resonance imaging (MRI) play a crucial role in accurately assessing glioblastoma. While these imaging modalities have greatly improved diagnosis, the identification of genetic abnormalities for targeted therapy often necessitates invasive procedures, carrying significant risks\u0026nbsp;[14-16]. A promising solution to these challenges lies in radiogenomics, a non-invasive approach that analyzes radiomic features extracted from various imaging modalities to detect and identify tumors. Radiogenomics combines genetics, radiomics, and artificial intelligence (AI), offering new possibilities for targeted therapy and precision medicine by enabling the identification of genetic profiles. By analyzing phenotypic data obtained from medical imaging, radiogenomics provides insights into tumor behavior and treatment response\u0026nbsp;[17-20]. Recent studies have shown direct associations between tumors\u0026apos; phenotypic and genotypic characteristics, sparking interest in extracting genotypic information from medical imaging techniques\u0026nbsp;[21-23].\u003c/p\u003e\n\u003cp\u003eThe disruption of signaling pathways such as Ras-ERK (extracellular signal-regulated kinase) and PI3K (phosphatidylinositol 3-kinase) can lead to abnormal cell behavior and contribute to cancer progression [24-26]. Targeted inhibition of these pathways using precision drugs offers a personalized treatment strategy [27, 28]. Glioblastoma, a formidable opponent in cancer research, effectively utilizes various signaling pathways to fuel the continuous growth and survival of cancerous cells. This coordinated network plays a crucial role in sustaining the aggressive behavior of the tumor. One of the main pathways is PI3K, which plays a vital role in glioblastoma\u0026apos;s pathophysiology. The PI3K pathway often becomes overactive due to the loss of the PTEN ( Phosphatase and Tensin Homolog) tumor suppressor gene, promoting tumor growth [29]. Additionally, NOTCH (Neurogenic Locus Notch Homolog Protein) signaling, which is essential in gliomagenesis, interacts with the TP53 pathway, affecting cell death [30-33]. Unfortunately, TP53 (Tumor protein 53), a vital tumor suppressor, is frequently disrupted in GBM [29]. Furthermore, The Receptor tyrosine kinase (RTK-RAS) pathway, commonly altered in glioblastoma, is linked to overactive RTK receptors, further fueling the complexity of the disease [34-36]. Moreover, WNT signaling, which is essential for CNS development, becomes dysregulated in GBM, especially in glioblastoma stem cells (GSCs), which contributes to tumor growth and makes it resistant to drugs [37-40]. Traditionally, the identification of these pathways relies upon gene expression analysis, a laborious and resource-intensive process [41]. However, the advent of radiomics holds immense promise in revolutionizing GBM management. By utilizing imaging modalities like MRI, radiomics facilitates non-invasive elucidation of underlying molecular processes, potentially offering personalized treatment strategies and substantially enhancing patient outcomes. In this context, our study explores the utility of radiogenomics and machine learning (ML) in predicting oncogenic signaling pathways in GBM. By extracting radiomic features from MRI scans and integrating them with genomic data, we aim to identify associations between imaging phenotypes and genetic profiles, mainly focusing on key signaling pathways implicated in gliomagenesis. Through advanced ML algorithms, we seek to develop predictive models that accurately identify these pathways, thereby informing personalized therapeutic approaches and improving patient outcomes.\u003c/p\u003e"},{"header":"Methodology","content":"\u003ch2\u003eData Collection of MRI Scans and Signalling Pathways\u003c/h2\u003e\n\u003cp\u003eOur experimental setup commenced by acquiring post-operative scans with multiple parameters from various modalities, comprising T1-weighted (T1w), contrast-enhanced T1-weighted (T1c), fluid-attenuated inversion recovery (FLAIR), and T2-weighted (T2w) images. These scans were sourced from the BRATS-19 dataset, made available by the Centre for Biomedical Image Computing \u0026amp; Analytics (CBCIA) at the University of Pennsylvania\u0026nbsp;[42-44]. The dataset comprised MRI scans obtained from patients diagnosed with both GBM and Low-Grade Glioma (LGG), accompanied by publicly accessible genomic and other clinical data, which can be accessed through platforms such as The Cancer Genome Atlas (TCGA)\u0026nbsp;[35]\u0026nbsp;and Clinical Proteomic Tumour Analysis Consortium (CPTAC)\u0026nbsp;[45]. CBCIA offers a file-name mapping that correlates the provided scans with patient identifiers in TCGA and CPTAC portals, facilitating access to genetic and clinical information from alternative sources. The signaling pathways of the relevant patients were gathered from cBioPortal. This platform offers interactive access to genomic profiles across different datasets and hosts datasets of signaling pathways associated with these profiles. Due to the absence of an API for accessing the pathway dataset, the pathway data was manually extracted using a web scraper. This extraction was performed from the GBM TCGA PanCancer Atlas (Study ID: gbm_tcga_pan_can_atlas_2018), GBM CPTAC (Study ID: gbm_cptac_2021), and Brain Lower Grade Glioma TCGA PanCancer (Study ID: gbm_tcga_pan_can_atlas_2018) datasets. In cases where certain pathways were unavailable in these datasets, they were obtained alternatively from the TCGA Firehose Legacy dataset (Study ID: gbm_tcga, lgg_tcga).\u003c/p\u003e\n\u003cp\u003eWhile the training set encompassed MRI scans of LGG, it's noteworthy that GBM and LGG exhibit unique features and certain discrepancies in appearance. However, despite these disparities, the two types of gliomas also display overlapping imaging features and share specific genetic characteristics. Both tumors show increased signal intensity in T2-weighted scans, decreased intensity at T1-weighted imaging, and exhibit contrast enhancement when a contrasting agent is used\u0026nbsp;[46]. The MRI Scans in both cases show infiltrative growth patterns by exhibiting abnormal signal intensity beyond the tumor mass to indicate the spread of the tumor\u0026nbsp;[47]. They show similarities in genetic mutation\u0026nbsp;[48, 49]\u0026nbsp;and also have similarities in the aberrations of their signaling pathways\u0026nbsp;[50, 51].\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eRadiomic Feature Extraction of MRI Scans\u003c/h2\u003e\n\u003cp\u003eA panel of radiomic features was extracted from the four MRI modalities using the segmentation mask provided for each subject. The extraction process was conducted utilizing PyRadiomics\u0026nbsp;[52], which includes feature definitions that are compliant with the Imaging Biomarker Standardization Initiative (IBSI)\u0026nbsp;[53]. The IBSI standardizes feature definitions and furnishes reference values for verifying radiomic software, enhancing reproducibility, and facilitating the clinical translation of radiomic research. The extracted feature sets underwent standardization to achieve a normal distribution and normalization within the range [0,1]. However, the abundance of extracted features, numbering over a thousand (\u0026gt;1000), could potentially lead to the \"curse of dimensionality.\"\u0026nbsp;[54]. Dimensionality reduction and selection of the most relevant features for the outcome model were performed through feature engineering on the generated feature panel.\u003c/p\u003e\n\u003ch2\u003eData Balancing\u003c/h2\u003e\n\u003cp\u003eTo address the data imbalance issue in our dataset, we employed the Synthetic Minority Oversampling Technique (SMOTE)\u0026nbsp;[55]. SMOTE utilizes the K-Nearest Neighbors algorithm to identify neighboring feature vectors within the minority class in a given vector space. Subsequently, it generates synthetic data points along the line connecting these neighboring feature vectors of the minority class. This results in an expanded and less narrowly defined decision boundary between the classes, consequently enhancing the performance of classification models. The newly created instance stays within the original feature space, maintaining the dataset's characteristics and distribution. Furthermore, the generated data introduces extra decision boundaries for a more generalized model, unlike methods such as random oversampling, which merely replicates existing instances.\u003c/p\u003e\n\u003ch2\u003eModel Selection and Training\u003c/h2\u003e\n\u003cp\u003eFive supervised classification models based on ML were chosen and trained using the provided feature set to forecast the five signaling pathways. The selected algorithms comprise a Logistic Regression Classifier (LRC), Support Vector Machine (SVM), Random Forest Classifier (RFC), AdaBoost Classifier (ABC), and K-Nearest Neighbor Classifier (KNN). To enhance the prediction accuracy of our models, we employed Grid Search for hyperparameter tuning, aiming to determine the optimal hyperparameter configurations for each algorithm. The models underwent 5-fold cross-validation to ensure a more precise and unbiased performance assessment. Subsequently, the models with the most suitable hyperparameter settings were evaluated on separate, unseen test sets. Various evaluation metrics were used to identify each signaling pathway's top-performing algorithms.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eWe examine the results of our experiments through cross-validation using four classification algorithms: RFC, SVC, ABC, and LRC. The objective was to detect five specific signaling pathways (WNT, PI3K, TP53, RTK-RAS, NOTCH) across three different datasets: over_sample, under_sample, and under_sample_pure. Our approach involved employing a 5-fold cross-validation technique and evaluating the algorithms' performance based on accuracy, precision, recall, and F1-score.\u003c/p\u003e\n\u003cp\u003eMRI Scans and Segmentation Labels\u003c/p\u003e\n\u003cp\u003eThe multi-institutional pre-operative MRI scans for GBM and LGG patients of TCGA (n =167) and CPTAC (n =19)\u0026nbsp;were made available in the BRATS-19 dataset released by CBCIA. In addition to containing T1, T2, Post Contrast T1, and FLAIR 3D MRI volumes, the dataset comprises segmentation labels experienced neuro-radiologists have verified. These segmentation labels include annotations for the Enhancing Tumor part (ET), Necrotic and non-enhancing Tumor part (NET), and Peritumoral Edema (ED). The scans have undergone various preprocessing steps, including skull-stripping, co-registration, and interpolation, to achieve a resolution of 1 mm³.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCollection of Signaling Pathways and Mapping with MRI Scans\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNine oncogenic signaling pathways listed in \u003cstrong\u003eTable 1\u003c/strong\u003e of the feature set were extracted from cBioPortal. However, two pathways, NRF2 and TGF-β, exhibited no alterations and were consequently excluded from further analysis. As depicted in \u003cstrong\u003eFigure 2\u003c/strong\u003e, the distribution of pathway alterations is imbalanced, with either an excess or a scarcity of alterations observed. In situations with a significant disparity between the majority and minority classes, ML algorithms tend to bias their classification outcomes towards the majority class, leading to skewed results\u0026nbsp;[56]. While achieving high accuracy, the algorithm may not perform optimally regarding other performance metrics like sensitivity (recall) or F1-Score. Despite its strong performance, SMOTE may not effectively handle severe imbalances in the data. To tackle this challenge, the top four oncogenic signaling pathways with the least imbalance (\u0026lt;30%) were chosen for model training. These pathways include PI3K, TP53, RTK-RAS, WNT, and NOTCH signaling pathways, which have been demonstrated to impact GBM significantly\u0026nbsp;[29-34, 36-41].\u003c/p\u003e\n\u003cp\u003eThe RTK-RAS, PI3K, NOTCH, TP53, and WNT signaling pathways are interconnected, with their interactions and cross-regulation playing a substantial role in the development and advancement of GBM\u0026nbsp;[29-34, 36-41]. Recognizing these signaling pathways and understanding the crosstalk between them is essential for administering and advancing targeted therapies. The relationships among signaling pathways across the subjects are depicted in an upset plot in \u003cstrong\u003eFigure 3\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eDerivation of Radiomic Feature Panel\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA comprehensive radiomic feature set of 1284 features, derived from 101 standard features following IBSI standards, was obtained from the four imaging sequences (T1, T2, T1c, FLAIR) along with their corresponding segmentation masks. The extracted features encompassed first-order, volumetric, and intensity-based textural features categorized as First Order, Shape, GLCM, GLDM, GLRLM, GLSZM, and NGTDM, with all 107 features listed in \u003cstrong\u003eTable 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eDimensionality Reduction and Feature Selection\u003c/p\u003e\n\u003cp\u003eTo diminish the dimensionality of the feature set, features exhibiting significant correlation with the target pathway (\u0026gt;0.1) and minimal correlation with each other (\u0026lt;0.9) were chosen from the radiomic feature panel. Subsequently, feature importance was assessed for each training set using the random forest algorithm in sklearn with default parameter settings to identify the top 10 most important features. The top 10 features selected for each pathway are presented in \u003cstrong\u003eTable 3\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eCreation of Datasets\u003c/p\u003e\n\u003cp\u003eAfter preprocessing, feature engineering, and data splitting, the final data cohort consists of 167 subjects from TCGA (GBM =98, LGG =69) and 19 from CPTAC. It would’ve been suitable to set the 19 subjects from CTPAC aside for data validation; however, due to heavy imbalance in the signaling pathways of the CPTAC dataset, the experimentation was carried out on three separate datasets created by splitting the cohort into training and validation sets (\u003cstrong\u003eFigure 4\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ei) In the over_split dataset, the validation set was created by combining samples from the CPTAC dataset with some samples from TCGA. The selection of TCGA samples depended on the availability of samples for the minority class in TCGA. There were specific criteria for constructing this validation set. If the minority class in the CPTAC dataset represented the majority class in TCGA-GBM, then all samples of the minority class from the training set were incorporated into the validation set. Similarly, if the minority class was consistent between TCGA-GBM and CPTAC, samples from TCGA-GBM were chosen for inclusion in the validation set. The number of cases transferred to the validation set depended on the ratio of the majority class to the minority class in TCGA-GBM. If this ratio was less than 0.11, three cases were moved to the validation set; otherwise, five cases were moved to ensure balanced representation. \u0026nbsp;This was done to balance the class distribution.\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"614\" height=\"162\"\u003e\u003c/p\u003e\n\u003cp\u003eii) The under_split dataset features a validation set comprising solely CPTAC samples, balanced using under-sampling techniques. This involved restricting the number of samples in the majority class to twice that of the minority class. The remaining samples from the majority class were then utilized to augment the size of the training set, thereby enhancing the training process.\u003c/p\u003e\n\u003cp\u003eiii) The under_split_pure dataset encompassed a validation set comprising exclusively CPTAC samples, which were balanced via under-sampling. This entailed restricting the number of samples in the majority class to twice that of the minority class, with any surplus samples being discarded.\u003c/p\u003e\n\u003ch2\u003eEvaluation of ML Models using K-Fold Cross Validation\u003c/h2\u003e\n\u003cp\u003eFive ML algorithms were trained to detect five fundamental oncogenic signaling pathways by training them on diverse radiomic features extracted from segmentation labels of GBM MRI scans. The models underwent training using five-fold cross-validation to assess the generalizability of the employed approach. The mean results of the 5-fold cross-validation of each model across all three datasets in terms of accuracy are tabulated in \u003cstrong\u003eTable 4\u003c/strong\u003e and further visualized for clarity in \u003cstrong\u003eFigure 5\u003c/strong\u003e. In addition to accuracy, the ROC_AUC score was selected to provide a more comprehensive assessment of the models' performance. The mean results here showcase the optimal parameters achieved through thorough hyperparameter tuning using Grid Search. The precise hyperparameter configuration responsible for these results is outlined in \u003cstrong\u003eTable 5\u003c/strong\u003e.\u003c/p\u003e\n\u003ch2\u003eModel Validation on Unseen Data\u003c/h2\u003e\n\u003cp\u003eThe model that performed the best on each signaling pathway was further validated on the test sets set aside to evaluate the generalizability of the models. The following metrics were chosen to give an idea of the general prediction power, showcase the performance on imbalanced data, and observe the trade-off between the correct prediction rate and misclassifications. Accuracy may not be the most suitable metric in classification problems when dealing with moderate to severe data imbalance. In such scenarios, precision, recall (specificity), and F1-score offer a more precise assessment of the models' performance. Comparative visualizations of the algorithms based on accuracy, precision, recall, and F1-score are presented in \u003cstrong\u003eFigure 6.\u003c/strong\u003e\u003c/p\u003e\n\n\n\n\n\n\n\n"},{"header":"Discussion and Conclusion","content":"\u003cp\u003eUnderstanding the interconnectedness of signaling pathways like RTK-RAS, PI3K, NOTCH, and TP53 is vital for comprehending how they influence GBM's development and advancement. Detecting these pathways and their mutual communication is pivotal for administering existing treatments and advancing targeted therapies. Interesting associations can also be seen in our data and shown in the upset plot in \u003cstrong\u003eFigure 3\u003c/strong\u003e. PI3K, RTK-RAS, or TP53 alterations occur independently in no more than three cases each, whereas the NOTCH pathway was observed separately in nine cases and the WNT pathway in 16 cases. Conversely, PI3K, RTK-RAS, and TP53 are often co-present, with various combinations observed. For instance, unique combinations of RTK-RAS and TP53 were found in 31 instances, TP53 with only PI3K in 14 cases, and all three together in 30 cases. These three pathways are well-known for gliomagensis\u0026nbsp;[35, 57, 58]. In addition to the simultaneous presence of established pathways, there are instances where pathways such as NOTCH, which is recognized for inducing apoptosis in the P53 gene, are found alongside the TP53 pathway in 35 cases. The interaction between these pathways is extensively studied and documented\u0026nbsp;[33, 59].\u0026nbsp;Notably, all four pathways are present in just seven out of the total 167 cases.\u003c/p\u003e\u003cp\u003eThe research highlights the significance of carefully selecting algorithms, ensuring data quality, and conducting thorough feature engineering in radio-genomic studies aimed at pathway detection. Each pathway demonstrates distinct interactions and presents specific challenges in this regard. Various ML models demonstrate varying evaluation scores, suggesting that particular pathways may present more complex classification tasks than others. For example, the PI3K pathways seemed more difficult for models to classify accurately, as evidenced by lower precision, recall, and F1-score values. Variations in class distribution also influence model performance, as different representations of classes introduce biases in learning. Specific pathways, like NOTCH and PI3K, seemed to pose more significant challenges for models than others, as reflected in lower precision, recall, and F1-score values. This suggests that the classification boundaries within these pathways are more complex.\u003c/p\u003e\u003cp\u003eThe varied performance observed among different ML algorithms corresponds with the distinct challenges presented by each pathway and the size of the dataset. Ensemble methods, notably Random Forest, demonstrated consistent performance across different situations, indicating their potential as dependable baseline models. The TP53 pathway, renowned for its function as a tumor suppressor, yielded exciting findings. When applied to the over_split dataset, the ABC algorithm showed remarkable accuracy, precision, and F1-score, indicating its efficacy in detecting this pathway. Yet, the algorithm's efficacy dramatically declined on the under_split and under_split_pure datasets, with most algorithms showing exceptionally low precision and recall, likely due to the minimal test set comprising only three sample points. Conversely, cross-validation accuracy on these two datasets remained consistently above 0.70, except for LRC.\u003c/p\u003e\u003cp\u003eThe RTK-RAS pathway, characterized by its intricate network of interactions, displayed diverse performance across different datasets. In the over_split dataset, the SVC exhibited a balanced performance. However, except KNN, all algorithms failed to identify any true negatives, classifying all nine samples as positives, leading to 100% recall but zero precision. This discrepancy contrasts with the cross-validation outcomes on the training set, suggesting that none of the models have overfit and have not achieved successful generalization.\u003c/p\u003e\u003cp\u003eThe PI3K pathway, which plays a crucial role in cell growth and survival, exhibited relatively consistent performance trends on under_split_pure and over_split datasets. In the case of over_split, the RFC achieved consistent results across all metrics, reflecting its robustness. However, on the under_split dataset, all algorithms faced challenges detecting any true negatives primarily due to possible class imbalance issues. Interestingly, the RFC algorithm excelled on the under_split_pure dataset, with all 9 cases correctly detected. The NOTCH signaling pathway is known for its significance in gliomagenesis. Across all datasets, we observed consistently higher performance among all the algorithms over all three datasets in the prediction of NOTCH Pathway. The LRC displayed exceptional precision and recall on the over_split and under_split datasets. In contrast, all algorithms failed to perform significantly on the under_split_pure dataset. \u0026nbsp;\u003c/p\u003e\u003cp\u003eThe WNT signaling pathway, critical for cell differentiation in the central nervous system, presented diverse performance across algorithms. We had the highest class imbalance and the least number of alterations among the pathways, which performed relatively poorly on all the datasets. RFC showed variable performance across datasets, while the LRC achieved high precision and recall on the under_split dataset. The SVC demonstrated strong performance on the under_split_pure dataset, indicating its capability to handle class imbalance effectively. This research revealed insights into utilizing machine learning models with radiomic data to forecast specific oncogenic signaling pathways. The findings underscore the impact of dataset size, class distribution, and feature complexity on model effectiveness. By considering these elements, we can enhance our prediction algorithms, fostering a deeper understanding of employing AI in radiomics to elucidate the interactions among different signaling pathways and their influence on tumor phenotypic traits.\u0026nbsp;\u003c/p\u003e\u003cp\u003eThe prediction of oncogenic signaling pathways from radiomic features holds promise for advancing genomic diagnosis faster and more cost-effectively. Invasive diagnostic procedures for brain tumors, such as brain biopsies, entail additional risks, making the timely and accurate genetic profiling of specimens crucial for targeted therapeutic interventions in Glioblastoma cases. Our study deployed four machine-learning models to forecast four oncogenic signaling pathways using MRI scans from the TCGA-GBM dataset. Our findings revealed a positive correlation between the radiomic features extracted from MRI scans and oncogenic signaling pathways in GBM. With adequate data, manual feature extraction could be bypassed, leading to the development of a more generalized multi-label deep learning model capable of predicting additional signaling pathways. We intend to expand this research by developing a multi-label deep learning model that can predict a broader spectrum of signaling pathways.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot Applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot Applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of supporting data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot Applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthor MK was employed by DigiBiomics Inc. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was funded, in part, by a Research Grant (Grant number: ID No. 2022-16465) from the Indian Council of Medical Research (ICMR) Govt. of India, New Delhi to Muzafar A. Macha. Promotion of University Research and Scientific Excellence (PURSE) (SR/PURSE/2022/121) grant from the Department of Science and Technology, Govt. of India, New Delhi to the Islamic University of Science and Technology (IUST), Awantipora.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eABA, SWA, MRB, and MAM wrote the manuscript and generated figures. TAM, AA, MAM, and MRB contributed to the concept \u0026amp; design and critically edited the manuscript. ABA, SWA and TAM performed experiments. ABA, SWA, TAM, AS, MAK, AAB, AS, MRB, and MAM critically revised and edited the scientific content. All authors read and approved the final manuscript.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGrochans S, Cybulska AM, Simińska D, Korbecki J, Kojder K, Chlubek D, Baranowska-Bosiacka I: \u003cstrong\u003eEpidemiology of glioblastoma multiforme\u0026ndash;literature review\u003c/strong\u003e. \u003cem\u003eCancers \u003c/em\u003e2022, \u003cstrong\u003e14\u003c/strong\u003e(10):2412.\u003c/li\u003e\n\u003cli\u003eLu I-N, Dobersalske C, Rauschenbach L, Teuber-Hanselmann S, Steinbach A, Ullrich V, Prasad S, Blau T, Kebir S, Siveke JT: \u003cstrong\u003eTumor-associated hematopoietic stem and progenitor cells positively linked to glioblastoma progression\u003c/strong\u003e. \u003cem\u003eNature communications \u003c/em\u003e2021, \u003cstrong\u003e12\u003c/strong\u003e(1):3895.\u003c/li\u003e\n\u003cli\u003eKim HJ, Park JW, Lee JH: \u003cstrong\u003eGenetic architectures and cell-of-origin in glioblastoma\u003c/strong\u003e. \u003cem\u003eFrontiers in oncology \u003c/em\u003e2021, \u003cstrong\u003e10\u003c/strong\u003e:615400.\u003c/li\u003e\n\u003cli\u003eMelhem JM, Detsky J, Lim-Fat MJ, Perry JR: \u003cstrong\u003eUpdates in IDH-wildtype glioblastoma\u003c/strong\u003e. \u003cem\u003eNeurotherapeutics \u003c/em\u003e2022, \u003cstrong\u003e19\u003c/strong\u003e(6):1705-1723.\u003c/li\u003e\n\u003cli\u003eTykocki T, Eltayeb M: \u003cstrong\u003eTen-year survival in glioblastoma. A systematic review\u003c/strong\u003e. \u003cem\u003eJournal of Clinical Neuroscience \u003c/em\u003e2018, \u003cstrong\u003e54\u003c/strong\u003e:7-13.\u003c/li\u003e\n\u003cli\u003ePatel NP, Lyon KA, Huang JH: \u003cstrong\u003eThe effect of race on the prognosis of the glioblastoma patient: a brief review\u003c/strong\u003e. \u003cem\u003eNeurological research \u003c/em\u003e2019, \u003cstrong\u003e41\u003c/strong\u003e(11):967-971.\u003c/li\u003e\n\u003cli\u003eBai J, Varghese J, Jain R: \u003cstrong\u003eAdult glioma WHO classification update, genomics, and imaging: what the radiologists need to know\u003c/strong\u003e. \u003cem\u003eTopics in Magnetic Resonance Imaging \u003c/em\u003e2020, \u003cstrong\u003e29\u003c/strong\u003e(2):71-82.\u003c/li\u003e\n\u003cli\u003eLouis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, Hawkins C, Ng H, Pfister SM, Reifenberger G: \u003cstrong\u003eThe 2021 WHO classification of tumors of the central nervous system: a summary\u003c/strong\u003e. \u003cem\u003eNeuro-oncology \u003c/em\u003e2021, \u003cstrong\u003e23\u003c/strong\u003e(8):1231-1251.\u003c/li\u003e\n\u003cli\u003eMcNeill KA: \u003cstrong\u003eEpidemiology of brain tumors\u003c/strong\u003e. \u003cem\u003eNeurologic clinics \u003c/em\u003e2016, \u003cstrong\u003e34\u003c/strong\u003e(4):981-998.\u003c/li\u003e\n\u003cli\u003eOstrom QT, Gittleman H, Farah P, Ondracek A, Chen Y, Wolinsky Y, Stroup NE, Kruchko C, Barnholtz-Sloan JS: \u003cstrong\u003eCBTRUS statistical report: Primary brain and central nervous system tumors diagnosed in the United States in 2006-2010\u003c/strong\u003e. \u003cem\u003eNeuro-oncology \u003c/em\u003e2013, \u003cstrong\u003e15\u003c/strong\u003e(suppl_2):ii1-ii56.\u003c/li\u003e\n\u003cli\u003eChakrabarti I, Cockburn M, Cozen W, Wang YP, Preston‐Martin S: \u003cstrong\u003eA population\u003c/strong\u003e\u003cstrong\u003e‐based description of glioblastoma multiforme in Los Angeles County, 1974\u0026ndash;1999\u003c/strong\u003e. \u003cem\u003eCancer: Interdisciplinary International Journal of the American Cancer Society \u003c/em\u003e2005, \u003cstrong\u003e104\u003c/strong\u003e(12):2798-2806.\u003c/li\u003e\n\u003cli\u003eOstrom QT, Gittleman H, Truitt G, Boscia A, Kruchko C, Barnholtz-Sloan JS: \u003cstrong\u003eCBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2011\u0026ndash;2015\u003c/strong\u003e. \u003cem\u003eNeuro-oncology \u003c/em\u003e2018, \u003cstrong\u003e20\u003c/strong\u003e(suppl_4):iv1-iv86.\u003c/li\u003e\n\u003cli\u003eOstrom QT, Price M, Neff C, Cioffi G, Waite KA, Kruchko C, Barnholtz-Sloan JS: \u003cstrong\u003eCBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2015-2019\u003c/strong\u003e. \u003cem\u003eNeuro Oncol \u003c/em\u003e2022, \u003cstrong\u003e24\u003c/strong\u003e(Suppl 5):v1-v95.\u003c/li\u003e\n\u003cli\u003eBi WL, Hosny A, Schabath MB, Giger ML, Birkbak NJ, Mehrtash A, Allison T, Arnaout O, Abbosh C, Dunn IF: \u003cstrong\u003eArtificial intelligence in cancer imaging: clinical challenges and applications\u003c/strong\u003e. \u003cem\u003eCA: a cancer journal for clinicians \u003c/em\u003e2019, \u003cstrong\u003e69\u003c/strong\u003e(2):127-157.\u003c/li\u003e\n\u003cli\u003eHussain T, Nguyen QT: \u003cstrong\u003eMolecular imaging for cancer diagnosis and surgery\u003c/strong\u003e. \u003cem\u003eAdvanced drug delivery reviews \u003c/em\u003e2014, \u003cstrong\u003e66\u003c/strong\u003e:90-100.\u003c/li\u003e\n\u003cli\u003eChen C-C, Hsu P-W, Wu T-WE, Lee S-T, Chang C-N, Wei K-c, Chuang C-C, Wu C-T, Lui T-N, Hsu Y-H: \u003cstrong\u003eStereotactic brain biopsy: Single center retrospective analysis of complications\u003c/strong\u003e. \u003cem\u003eClinical neurology and neurosurgery \u003c/em\u003e2009, \u003cstrong\u003e111\u003c/strong\u003e(10):835-839.\u003c/li\u003e\n\u003cli\u003eShui L, Ren H, Yang X, Li J, Chen Z, Yi C, Zhu H, Shui P: \u003cstrong\u003eThe era of radiogenomics in precision medicine: an emerging approach to support diagnosis, treatment decisions, and prognostication in oncology\u003c/strong\u003e. \u003cem\u003eFrontiers in Oncology \u003c/em\u003e2021, \u003cstrong\u003e10\u003c/strong\u003e:570465.\u003c/li\u003e\n\u003cli\u003eBodalal Z, Trebeschi S, Nguyen-Kim TDL, Schats W, Beets-Tan R: \u003cstrong\u003eRadiogenomics: bridging imaging and genomics\u003c/strong\u003e. \u003cem\u003eAbdominal radiology \u003c/em\u003e2019, \u003cstrong\u003e44\u003c/strong\u003e(6):1960-1984.\u003c/li\u003e\n\u003cli\u003eSundaram M, McGuire MH, Herbold DR: \u003cstrong\u003eMagnetic resonance imaging of soft tissue masses: an evaluation of fifty-three histologically proven tumors\u003c/strong\u003e. \u003cem\u003eMagnetic resonance imaging \u003c/em\u003e1988, \u003cstrong\u003e6\u003c/strong\u003e(3):237-248.\u003c/li\u003e\n\u003cli\u003eUpadhyay N, Waldman A: \u003cstrong\u003eConventional MRI evaluation of gliomas\u003c/strong\u003e. \u003cem\u003eThe British journal of radiology \u003c/em\u003e2011, \u003cstrong\u003e84\u003c/strong\u003e(special_issue_2):S107-S111.\u003c/li\u003e\n\u003cli\u003eZinn PO, Singh SK, Kotrotsou A, Hassan I, Thomas G, Luedi MM, Elakkad A, Elshafeey N, Idris T, Mosley J: \u003cstrong\u003eA coclinical radiogenomic validation study: conserved magnetic resonance radiomic appearance of periostin-expressing glioblastoma in patients and xenograft models\u003c/strong\u003e. \u003cem\u003eClinical Cancer Research \u003c/em\u003e2018, \u003cstrong\u003e24\u003c/strong\u003e(24):6288-6299.\u003c/li\u003e\n\u003cli\u003ePei L, Vidyaratne L, Rahman MM, Iftekharuddin KM: \u003cstrong\u003eContext aware deep learning for brain tumor segmentation, subtype classification, and survival prediction using radiology images\u003c/strong\u003e. \u003cem\u003eScientific Reports \u003c/em\u003e2020, \u003cstrong\u003e10\u003c/strong\u003e(1):19726.\u003c/li\u003e\n\u003cli\u003eLi K, Xiao J, Yang J, Li M, Xiong X, Nian Y, Qiao L, Wang H, Eresen A, Zhang Z: \u003cstrong\u003eAssociation of radiomic imaging features and gene expression profile as prognostic factors in pancreatic ductal adenocarcinoma\u003c/strong\u003e. \u003cem\u003eAmerican journal of translational research \u003c/em\u003e2019, \u003cstrong\u003e11\u003c/strong\u003e(7):4491.\u003c/li\u003e\n\u003cli\u003eSever R, Brugge JS: \u003cstrong\u003eSignal transduction in cancer\u003c/strong\u003e. \u003cem\u003eCold Spring Harbor perspectives in medicine \u003c/em\u003e2015, \u003cstrong\u003e5\u003c/strong\u003e(4):a006098.\u003c/li\u003e\n\u003cli\u003eVillanueva A, Garc\u0026iacute;a C, Paules AB, Vicente M, Meg\u0026iacute;as M, Reyes G, de Villalonga P, Agell N, Llu\u0026iacute;s F, Bachs O: \u003cstrong\u003eDisruption of the antiproliferative TGF-\u0026beta; signaling pathways in human pancreatic cancer cells\u003c/strong\u003e. \u003cem\u003eOncogene \u003c/em\u003e1998, \u003cstrong\u003e17\u003c/strong\u003e(15):1969-1978.\u003c/li\u003e\n\u003cli\u003eMendoza MC, Er EE, Blenis J: \u003cstrong\u003eThe Ras-ERK and PI3K-mTOR pathways: cross-talk and compensation\u003c/strong\u003e. \u003cem\u003eTrends in biochemical sciences \u003c/em\u003e2011, \u003cstrong\u003e36\u003c/strong\u003e(6):320-328.\u003c/li\u003e\n\u003cli\u003eSamimi H, Fallah P, Sohi AN, Tavakoli R, Naderi M, Soleimani M, Larijani B, Haghpanah V: \u003cstrong\u003ePrecision medicine approach to anaplastic thyroid cancer: advances in targeted drug therapy based on specific signaling pathways\u003c/strong\u003e. \u003cem\u003eActa Medica Iranica \u003c/em\u003e2017:200-208.\u003c/li\u003e\n\u003cli\u003eZhu P, Aliabadi HM, Uludağ H, Han J: \u003cstrong\u003eIdentification of potential drug targets in cancer signaling pathways using stochastic logical models\u003c/strong\u003e. \u003cem\u003eScientific reports \u003c/em\u003e2016, \u003cstrong\u003e6\u003c/strong\u003e(1):23078.\u003c/li\u003e\n\u003cli\u003eMischel PS, Nelson SF, Cloughesy TF: \u003cstrong\u003eMolecular analysis of glioblastoma: pathway profiling and its implications for patient therapy\u003c/strong\u003e. \u003cem\u003eCancer biology \u0026amp; therapy \u003c/em\u003e2003, \u003cstrong\u003e2\u003c/strong\u003e(3):242-247.\u003c/li\u003e\n\u003cli\u003ePearson JR, Regad T: \u003cstrong\u003eTargeting cellular pathways in glioblastoma multiforme\u003c/strong\u003e. \u003cem\u003eSignal transduction and targeted therapy \u003c/em\u003e2017, \u003cstrong\u003e2\u003c/strong\u003e(1):1-11.\u003c/li\u003e\n\u003cli\u003eBazzoni R, Bentivegna A: \u003cstrong\u003eRole of notch signaling pathway in glioblastoma pathogenesis\u003c/strong\u003e. \u003cem\u003eCancers \u003c/em\u003e2019, \u003cstrong\u003e11\u003c/strong\u003e(3):292.\u003c/li\u003e\n\u003cli\u003eGaiano N, Fishell G: \u003cstrong\u003eThe role of notch in promoting glial and neural stem cell fates\u003c/strong\u003e. \u003cem\u003eAnnual review of neuroscience \u003c/em\u003e2002, \u003cstrong\u003e25\u003c/strong\u003e(1):471-490.\u003c/li\u003e\n\u003cli\u003eYang X, Klein R, Tian X, Cheng H-T, Kopan R, Shen J: \u003cstrong\u003eNotch activation induces apoptosis in neural progenitor cells through a p53-dependent pathway\u003c/strong\u003e. \u003cem\u003eDevelopmental biology \u003c/em\u003e2004, \u003cstrong\u003e269\u003c/strong\u003e(1):81-94.\u003c/li\u003e\n\u003cli\u003eSanchez-Vega F, Mina M, Armenia J, Chatila WK, Luna A, La KC, Dimitriadoy S, Liu DL, Kantheti HS, Saghafinia S: \u003cstrong\u003eOncogenic signaling pathways in the cancer genome atlas\u003c/strong\u003e. \u003cem\u003eCell \u003c/em\u003e2018, \u003cstrong\u003e173\u003c/strong\u003e(2):321-337. e310.\u003c/li\u003e\n\u003cli\u003eCGARNTssDUMSMRFABD, 5 EUVMEGBDJMMGOJJ, 8 HFHMTLN, 11 MACCAKAYWBO, 13 UoCSFVSBMPM: \u003cstrong\u003eComprehensive genomic characterization defines human glioblastoma genes and core pathways\u003c/strong\u003e. \u003cem\u003eNature \u003c/em\u003e2008, \u003cstrong\u003e455\u003c/strong\u003e(7216):1061-1068.\u003c/li\u003e\n\u003cli\u003eFonseca COd, Linden R, Futuro D, Gattass CR, Quirico-Santos T: \u003cstrong\u003eRas pathway activation in gliomas: a strategic target for intranasal administration of perillyl alcohol\u003c/strong\u003e. \u003cem\u003eArchivum immunologiae et therapiae experimentalis \u003c/em\u003e2008, \u003cstrong\u003e56\u003c/strong\u003e:267-276.\u003c/li\u003e\n\u003cli\u003eLee Y, Lee J-K, Ahn SH, Lee J, Nam D-H: \u003cstrong\u003eWNT signaling in glioblastoma and therapeutic opportunities\u003c/strong\u003e. \u003cem\u003eLaboratory investigation \u003c/em\u003e2016, \u003cstrong\u003e96\u003c/strong\u003e(2):137-150.\u003c/li\u003e\n\u003cli\u003eKomiya Y, Habas R: \u003cstrong\u003eWnt signal transduction pathways\u003c/strong\u003e. \u003cem\u003eOrganogenesis \u003c/em\u003e2008, \u003cstrong\u003e4\u003c/strong\u003e(2):68-75.\u003c/li\u003e\n\u003cli\u003eGuan R, Zhang X, Guo M: \u003cstrong\u003eGlioblastoma stem cells and Wnt signaling pathway: Molecular mechanisms and therapeutic targets\u003c/strong\u003e. \u003cem\u003eChinese Neurosurgical Journal \u003c/em\u003e2020, \u003cstrong\u003e6\u003c/strong\u003e:1-6.\u003c/li\u003e\n\u003cli\u003eZuccarini M, Giuliani P, Ziberi S, Carluccio M, Di Iorio P, Caciagli F, Ciccarelli R: \u003cstrong\u003eThe role of Wnt signal in glioblastoma development and progression: a possible new pharmacological target for the therapy of this tumor\u003c/strong\u003e. \u003cem\u003eGenes \u003c/em\u003e2018, \u003cstrong\u003e9\u003c/strong\u003e(2):105.\u003c/li\u003e\n\u003cli\u003eKabir MH, Patrick R, Ho JW, O\u0026rsquo;Connor MD: \u003cstrong\u003eIdentification of active signaling pathways by integrating gene expression and protein interaction data\u003c/strong\u003e. \u003cem\u003eBMC systems biology \u003c/em\u003e2018, \u003cstrong\u003e12\u003c/strong\u003e:77-87.\u003c/li\u003e\n\u003cli\u003eMenze BH, Jakab A, Bauer S, Kalpathy-Cramer J, Farahani K, Kirby J, Burren Y, Porz N, Slotboom J, Wiest R: \u003cstrong\u003eThe multimodal brain tumor image segmentation benchmark (BRATS)\u003c/strong\u003e. \u003cem\u003eIEEE transactions on medical imaging \u003c/em\u003e2014, \u003cstrong\u003e34\u003c/strong\u003e(10):1993-2024.\u003c/li\u003e\n\u003cli\u003eBakas S, Akbari H, Sotiras A, Bilello M, Rozycki M, Kirby JS, Freymann JB, Farahani K, Davatzikos C: \u003cstrong\u003eAdvancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features\u003c/strong\u003e. \u003cem\u003eScientific data \u003c/em\u003e2017, \u003cstrong\u003e4\u003c/strong\u003e(1):1-13.\u003c/li\u003e\n\u003cli\u003eBakas S, Reyes M, Jakab A, Bauer S, Rempfler M, Crimi A, Shinohara RT, Berger C, Ha SM, Rozycki M: \u003cstrong\u003eIdentifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge\u003c/strong\u003e. \u003cem\u003earXiv preprint arXiv:181102629 \u003c/em\u003e2018.\u003c/li\u003e\n\u003cli\u003eWhiteaker JR, Halusa GN, Hoofnagle AN, Sharma V, MacLean B, Yan P, Wrobel JA, Kennedy J, Mani D, Zimmerman LJ: \u003cstrong\u003eCPTAC Assay Portal: a repository of targeted proteomic assays\u003c/strong\u003e. \u003cem\u003eNature methods \u003c/em\u003e2014, \u003cstrong\u003e11\u003c/strong\u003e(7):703-704.\u003c/li\u003e\n\u003cli\u003eHaydar N, Alyousef K, Alanan U, Issa R, Baddour F, Al-Shehabi Z, Al-Janabi MH: \u003cstrong\u003eRole of Magnetic Resonance Imaging (MRI) in grading gliomas comparable with pathology: A cross-sectional study from Syria\u003c/strong\u003e. \u003cem\u003eAnnals of Medicine and Surgery \u003c/em\u003e2022, \u003cstrong\u003e82\u003c/strong\u003e:104679.\u003c/li\u003e\n\u003cli\u003eGuillevin R, Herpe G, Verdier M, Guillevin C: \u003cstrong\u003eLow-grade gliomas: The challenges of imaging\u003c/strong\u003e. \u003cem\u003eDiagnostic and Interventional imaging \u003c/em\u003e2014, \u003cstrong\u003e95\u003c/strong\u003e(10):957-963.\u003c/li\u003e\n\u003cli\u003eAoki K, Nakamura H, Suzuki H, Matsuo K, Kataoka K, Shimamura T, Motomura K, Ohka F, Shiina S, Yamamoto T: \u003cstrong\u003ePrognostic relevance of genetic alterations in diffuse lower-grade gliomas\u003c/strong\u003e. \u003cem\u003eNeuro-oncology \u003c/em\u003e2018, \u003cstrong\u003e20\u003c/strong\u003e(1):66-77.\u003c/li\u003e\n\u003cli\u003ePaugh BS, Qu C, Jones C, Liu Z, Adamowicz-Brice M, Zhang J, Bax DA, Coyle B, Barrow J, Hargrave D: \u003cstrong\u003eIntegrated molecular genetic profiling of pediatric high-grade gliomas reveals key differences with the adult disease\u003c/strong\u003e. \u003cem\u003eJournal of clinical oncology \u003c/em\u003e2010, \u003cstrong\u003e28\u003c/strong\u003e(18):3061.\u003c/li\u003e\n\u003cli\u003eNasrolahi A, Azizidoost S, Radoszkiewicz K, Najafi S, Ghaedrahmati F, Anbiyaee O, Khoshnam SE, Farzaneh M, Uddin S: \u003cstrong\u003eSignaling pathways governing glioma cancer stem cells behavior\u003c/strong\u003e. \u003cem\u003eCellular Signalling \u003c/em\u003e2023, \u003cstrong\u003e101\u003c/strong\u003e:110493.\u003c/li\u003e\n\u003cli\u003eNakada M, Kita D, Watanabe T, Hayashi Y, Teng L, Pyko IV, Hamada J-I: \u003cstrong\u003eAberrant signaling pathways in glioma\u003c/strong\u003e. \u003cem\u003eCancers \u003c/em\u003e2011, \u003cstrong\u003e3\u003c/strong\u003e(3):3242-3278.\u003c/li\u003e\n\u003cli\u003eVan Griethuysen JJ, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, Beets-Tan RG, Fillion-Robin J-C, Pieper S, Aerts HJ: \u003cstrong\u003eComputational radiomics system to decode the radiographic phenotype\u003c/strong\u003e. \u003cem\u003eCancer research \u003c/em\u003e2017, \u003cstrong\u003e77\u003c/strong\u003e(21):e104-e107.\u003c/li\u003e\n\u003cli\u003eZwanenburg A, Leger S, Valli\u0026egrave;res M, L\u0026ouml;ck S: \u003cstrong\u003eImage biomarker standardisation initiative\u003c/strong\u003e. \u003cem\u003earXiv preprint arXiv:161207003 \u003c/em\u003e2016.\u003c/li\u003e\n\u003cli\u003eLimkin EJ, Sun R, Dercle L, Zacharaki EI, Robert C, Reuz\u0026eacute; S, Schernberg A, Paragios N, Deutsch E, Fert\u0026eacute; C: \u003cstrong\u003ePromises and challenges for the implementation of computational medical imaging (radiomics) in oncology\u003c/strong\u003e. \u003cem\u003eAnnals of Oncology \u003c/em\u003e2017, \u003cstrong\u003e28\u003c/strong\u003e(6):1191-1206.\u003c/li\u003e\n\u003cli\u003eChawla NV, Bowyer KW, Hall LO, Kegelmeyer WP: \u003cstrong\u003eSMOTE: synthetic minority over-sampling technique\u003c/strong\u003e. \u003cem\u003eJournal of artificial intelligence research \u003c/em\u003e2002, \u003cstrong\u003e16\u003c/strong\u003e:321-357.\u003c/li\u003e\n\u003cli\u003eKrawczyk B: \u003cstrong\u003eLearning from imbalanced data: open challenges and future directions\u003c/strong\u003e. \u003cem\u003eProgress in Artificial Intelligence \u003c/em\u003e2016, \u003cstrong\u003e5\u003c/strong\u003e(4):221-232.\u003c/li\u003e\n\u003cli\u003eVenkatesan S, Lamfers ML, Dirven CM, Leenstra S: \u003cstrong\u003eGenetic biomarkers of drug response for small-molecule therapeutics targeting the RTK/Ras/PI3K, p53 or Rb pathway in glioblastoma\u003c/strong\u003e. \u003cem\u003eCNS oncology \u003c/em\u003e2016, \u003cstrong\u003e5\u003c/strong\u003e(2):77-90.\u003c/li\u003e\n\u003cli\u003eHan S, Wang P-F, Cai H-Q, Wan J-H, Li S-W, Lin Z-H, Yu C-J, Yan C-X: \u003cstrong\u003eAlterations in the RTK/Ras/PI3K/AKT pathway serve as potential biomarkers for immunotherapy outcome of diffuse gliomas\u003c/strong\u003e. \u003cem\u003eAging (Albany NY) \u003c/em\u003e2021, \u003cstrong\u003e13\u003c/strong\u003e(11):15444.\u003c/li\u003e\n\u003cli\u003eDotto GP: \u003cstrong\u003eCrosstalk of Notch with p53 and p63 in cancer growth control\u003c/strong\u003e. \u003cem\u003eNature reviews Cancer \u003c/em\u003e2009, \u003cstrong\u003e9\u003c/strong\u003e(8):587-595.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e\u003cstrong\u003e.\u0026nbsp;\u003c/strong\u003eSignaling Pathway Alterations\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eS. No\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePathway\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAlteration\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eTP53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e83.67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e\u003cem\u003e2\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003e\u003cem\u003eNRF2\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e\u003cem\u003e0.00%\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eNOTCH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e19.39%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eMYC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e6.12%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eWNT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e14.29%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eTGF-\u0026beta;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e0.00%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eRTK-RAS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e85.71%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003ePI3K\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e43.88%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.501%;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 57.115%;\"\u003e\n \u003cp\u003eHIPPO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31.384%;\"\u003e\n \u003cp\u003e11.22%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003cstrong\u003e. Feature List after Feature Selection\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"954\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 246px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNOTCH\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 255px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePI3K\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 217px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRTK-RAS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 236px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTP53\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-ET-T1Gd-SRE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-ET-T1Gd-GLN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-ED-T1Gd-Energy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eHISTO-ED-T2-Bin7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-NET-T2-Correlation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eSPATIAL-Parietal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLOBAL-ED-T1Gd-Kurtosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eINTENSITY-Mean-NET-FLAIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLSZM-ET-T2-LGZE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eVOLUME-NET-OVER-ED\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLOBAL-NET-T1-Variance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eHISTO-NET-FLAIR-Bin1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-NET-T1-Contrast\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eVOLUME-BRAIN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eHISTO-ET-T2-Bin2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-ED-T1Gd-Contrast\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-ET-T1-LRHGE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eTEXTURE-GLSZM-NET-T1Gd-SZE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-NGTDM-ED-T1Gd-Busyness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-NET-T1Gd-Correlation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eHISTO-ET-T1-Bin10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eHISTO-ET-FLAIR-Bin1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLOBAL-ET-T1-Skewness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eINTENSITY-Mean-ED-FLAIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eHISTO-NET-T1Gd-Bin5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-NET-FLAIR-HGRE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-ET-FLAIR-SRE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eTEXTURE-GLOBAL-ET-T1Gd-Skewness\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eHISTO-NET-FLAIR-Bin9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-NET-FLAIR-SumAverage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLSZM-ET-FLAIR-SZE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eHISTO-NET-T1-Bin7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-NET-T2-RLV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eTEXTURE-GLRLM-ET-T1Gd-SRE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLSZM-ET-T2-HGZE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eTEXTURE-NGTDM-ED-T1Gd-Busyness\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.7862%;\"\u003e\n \u003cp\u003eHISTO-ET-T2-Bin1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 26.7296%;\"\u003e\n \u003cp\u003eHISTO-ET-FLAIR-Bin9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.7463%;\"\u003e\n \u003cp\u003eTEXTURE-GLCM-ET-T1-Homogeneity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24.7379%;\"\u003e\n \u003cp\u003eHISTO-ET-T2-Bin4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3\u003c/strong\u003e\u003cstrong\u003e.\u0026nbsp;\u003c/strong\u003eHyper-parameter state\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"605\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAlgorithm\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHyperparameter\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePathways\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNOTCH\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePI3K\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRTK-RAS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTP53\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003eAdaBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003ealgorithm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eSAMME.R\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eSAMME.R\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eSAMME\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eSAMME\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003elearning_rate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003en_estimators\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1800\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1800\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003eLogistic Regression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003epenalty\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003el1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003el1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003el2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003el2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003esolver\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eliblinear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eliblinear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eliblinear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eliblinear\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003eRandom Forest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003ebootstrap\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eTRUE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eTRUE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eTRUE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eTRUE\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003ecriterion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003egini\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003egini\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003egini\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003egini\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003emax_features\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003en_estimators\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e1600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e2000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003eSupport Vector Machine\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003eC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003ekernel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003erbf\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003elinear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003elinear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003erbf\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Glioblastoma, Signalling Pathways, Radiogenomics, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-5131289/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5131289/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGlioblastoma (GBM) is a highly aggressive brain tumor associated with a poor patient prognosis. Despite standard therapies, the survival rate remains low, highlighting the urgent need for novel treatment strategies. Advanced imaging techniques, particularly magnetic resonance imaging (MRI), play a crucial role in assessing GBM. Disruptions in various oncogenic signalling pathways, such as Receptor Tyrosine Kinase (RTK)-Ras-Extracellular signal-regulated kinase (ERK) signalling, Phosphoinositide 3- Kinases (PI3Ks), tumour protein p53 (TP53), and Neurogenic locus notch homolog protein (NOTCH), contribute to the development of different tumour types, each exhibiting distinct morphological and phenotypic features that can be observed at a microscopic level. However, identifying genetic abnormalities for targeted therapy often requires invasive procedures, prompting exploration into non-invasive approaches like radiogenomics. This study explores the utility of radiogenomics and machine learning (ML) in predicting these oncogenic signaling pathways in GBM patients. Data from MRI scans and signaling pathways were collected, radiomic features were extracted, and ML models were trained and evaluated using cross-validation techniques. Our results showed a positive association between most signalling pathways and the radiomic features derived from MRI scans. The best models achieved high AUC scores, namely 0.7 for RTK-RAS, 0.8 for PI3K, 0.75 for TP53, and 0.4 for NOTCH, and therefore demonstrated the potential of ML models in accurately predicting oncogenic signaling pathways from radiomic features, thereby informing personalized therapeutic approaches and improving patient outcomes. We present a novel approach for the non-invasive prediction of deregulation in oncogenic signaling pathways in glioblastoma (GBM) by integrating radiogenomic data with machine learning (ML) models. \u0026nbsp;This research contributes to the advancement of precision medicine in GBM management, highlighting the importance of integrating radiomics with genomic data to better understand tumor behavior and treatment response.\u003c/p\u003e","manuscriptTitle":"Radiogenomics and Machine Learning Predict Oncogenic Signaling Pathways in Glioblastoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-19 14:31:37","doi":"10.21203/rs.3.rs-5131289/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d5a90699-ad44-4a9e-acfb-a19a641c7951","owner":[],"postedDate":"November 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-11-19T14:31:40+00:00","versionOfRecord":[],"versionCreatedAt":"2024-11-19 14:31:37","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5131289","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5131289","identity":"rs-5131289","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}