Fig. 3

Introduction of a machine learning-based genetic mutation analysis platform. A Schematic diagram showing an overview of DRIVE, a feature-based machine learning platform for pan-cancer assessment of somatic missense mutations, modified from Ref. [142]. This approach uses a total of 51 features spanning the gene and mutation levels. Several state-of-the-art supervised machine learning algorithms were applied to the final dataset, with results presented for the highest performing algorithms, including random forests, logistic regression, extreme gradient boosting, k-nearest neighbors, support vector machines, and multilayer perceptron. B Figure outlining a machine learning model for predicting response to immune checkpoint inhibitors, modified from Ref. [147]. Sixteen cancers were individually divided into training (80%) and testing (20%) subsets. To predict (responder and non-responder) immune checkpoint inhibitors, random forest models were trained on multiple genomic, molecular, demographic, and clinical characteristics on the training data using fivefold cross-validation. Consequently, trained models with optimal hyperparameters were evaluated on various performance measures using the test set