Table 2. Ligand-only CNN models that achieved high AUC (greater than 0.9) for COMT.

Test set	Model trained by	AUC	Sequence similarity*
COMT (Methyltransferase)	ADA (Adenosine deaminase)	0.934	11/21
	CASP3 (Caspase-3)	0.952	No Match
	CP3A4(Cytochrome P450)	0.901	8/23
	DEF (Peptide Deformylase)	0.950	No Match
	GRIA2(Glutamate receptor)	0.926	No Match
	HIVINT (HIV integrase)	0.998	3/8
	HMDH (HMG-CoA reductase)	0.930	11/33
	HS90A (Hot shock protein)	0.994	5/14
	INHA (Mycobacterium tuberculosis enoyl reductase)	0.964	8/32
	PPARG (Peroxisome proliferator-activated receptor)	0.951	15/54
	THB (Thyroid hormone receptor)	0.910	No Match

*From the NCBI BLASTp program using the default parameters. x/y is the sequence identity, where x is the number of identical amino acids in the local alignment and y is the total number of amino acids in the local alignment. “No Match” means no alignment was possible for the two sequences. All 11 targets are not homologous to COMT based on a 30% sequence similarity threshold.