Fig. 2From: binomialRF: interpretable combinatoric efficiency of random forests to identify biomarker interactionsBiomarker accuracies of the TCGA validation study. The TCGA validation study was conducted using breast and kidney cancer datasets, accessed via the R package TCGA2STAT. The matched-sample datasets were utilized to determine whether binomialRF could produce an accurate classifier via main effects and interactions. Left, the two binomialRF classifiers (51 identified gene main effects; 39 identified gene-gene interactions) and obtained a classifier as accurate as the original black-box RF model with all ~ 20,000 genes. Right, the two binomialRF classifiers (16 identified gene main effects; 11 identified gene-gene interactions) obtained a classifier as accurate as the original black-box RF model with all ~ 20,000 genesBack to article page