DeepAEG: a model for predicting cancer drug response based on data enhancement and edge-collaborative update strategies

Table 3 Top-5 prioritized cancer-associated genes identified by DeepAEG

Drug	Cell line	TCGA type	Top-5 related genes
Erlotinib	A3/KAW	DLBC	MTOR,ALK,RAF1,ETV5,EGFR
Lapatinib	HCC1599	BRCA	MDS2,FLCN ,EGFR ,VHL,FCGR2B
Crizotinib	GA-10	DLBC	ARHGEF10L,CDX2,CDG10,MET,CANT1
Dasatinib	NALM-6	ALL	IKBKB,THRAP3,CDKN1A,BCL2L12,SRC
Imatinib	KG-1	LAML	FOXR1,SFPQ,ERCC5,RAD17,ABL

We employed neural network automatic differentiation to compute the sum of gradients and rank the absolute value of each gene. Notably, certain top-5 genes have been corroborated by previous studies as being associated with drug responses or diseases corresponding to specific cell lines. These genes are indicated in bold

ISSN: 1471-2105