Table 2.

Summary of feature sets used in this study.

Name	Content	Reference
Gene-level feature sets
PTGS (all)	1,331 genes	Kohonen et al., 2017
PTGS (core)	199 genes	Kohonen et al., 2017
L1000	978 genes	Subramanian et al., 2017
Toxicity Module Gene	1,312 genes belonging to 89 gene co-expression modules for chemically induced liver injury	Tawa et al., 2014; Te et al., 2016
Toxicity Module (L1000)	154 genes common to Toxicity Module and L1000
A200	200 genes arbitrarily selected from the genome
A600	600 genes arbitrarily selected from the genome
A1200	1,200 genes arbitrarily selected from the genome
Pathway-/module-level feature sets
MSigDB (C2)	1,329 pathways in C2 collection
MSigDB (hallmark)	50 pathways in Hallmark collection	Liberzon et al., 2015
Toxicity Module	89 gene co-expression modules for chemically induced liver injuries	Tawa et al., 2014; Te et al., 2016
MSigDB (C2) L1000	1,220 MSigDB C2 pathways in which only genes in L1000 are retained
Y-shuffled feature sets
L1000 Y-Shuffle	L1000 data with random permutations of injury annotations
Toxicity Module Gene Y-Shuffle	Toxicity Module Gene data with random permutations of injury annotations