Fig. 1

An overall workflow of the proposed classifiers approach. Two datasets, TCGA TARGET GTEx and test set, were used in this study. After standardizing the two datasets, they were split for training, validation, and testing. Features were selected only from the training set and were used to build classifiers. To obtain classifiers with the highest performance, we employed Bayesian optimization for hyperparameter tuning. TCGA = The Cancer Genome Atlas; TARGET = tumor alterations relevant for genomics-driven therapy; GTEx = Genotype-Tissue Expression; FPKM = fragment per kilobase of transcript per million mapped reads; ENSGs = Ensembl genes; NDEG = not differentially expressed genes; MSS = mean sum of squares; ML = machine learning; DNN = deep feedforward neural network; SVM = support vector machine; RF = random forest; GB = gradient boosting