Fig. 5From: Considerations for feature selection using gene pairs and applications in large-scale dataset integration, novel oncogene discovery, and interpretable cancer screeningGene pairs facilitate large-scale dataset integration to identify BSDC1 and LRRC42 as novel breast cancer biomarkers. Transcriptomic profiles from 10,350 breast tumors were combined and the top 100 gene pairs were filtered using a TSP, b AM, c AA, d FM, and e FA. Connected genes represent a pair. f Genes within the top 100 pairs were often common to multiple feature selection methods. The top five most represented genes using g TSP, h AM, i AA, j FM, and k FA are reported. Y-axes represent the number of times the gene occurs in the top 100 pairsBack to article page