Table 1.

Appropriate metrics and number of nearest neighbors to assess the performance of various descriptor read-across prediction of chronic liver toxicity. Column 2 represents the descriptors type including: chemical structure (chm), biological (bio), and hybrid (CB). Descriptors names in column 3 include: Biological (B) − Assay (asy) and Gene (ge); Chemical Structure (C) − Morgan (mrgn), Toxprints (toxp), Topological Torsion (tptr), all chemicals combination descriptor (CC), Chemical + Biological Hybrid (CB) − Morgan + Assay (MA), Morgan + Gene (MG), Topological Torsion + Assay (TTA), Topological Torsion + Gene (TTG), Toxprints + Assay (TXA), Toxprints + Gene (TXG), and all chemical and biological descriptors combined (CB). Column 4 denotes the AUC performance values for each descriptor in predicting chronic liver toxicity effect. Column 5–6 denote the appropriate similarity metric and number of neighbors to make accurate prediction for each descriptor. Major descriptors of interest are marked bold.

Liver Effect	Descriptor Type	Descriptor Name	AUC	Metric	N Neighbors
Chr_liver	Chm	Tptr	0.6303	Euclidean	9
	Chm	Mrgn	0.64549	Jaccard	8
	Chm	Toxp	0.61379	Jaccard	7
	Bio	Ge	0.648847	Euclidean	14
	Bio	Asy	0.6632	Euclidean	11
	CB	mrgn + asy	0.6883	Jaccard	13
	CB	toxp + ge	0.7044	Jaccard	10
	CB	tptr + ge	0.6818	Euclidean	6
	CB	(CB) all	0.6999	Jaccard	14
	Chm	(CC) all	0.6702	Jaccard	10
	CB	mrgn + ge	0.7049	Jaccard	10
	CB	toxp + asy	0.6992	Jaccard	14
	CB	tptr + asy	0.6721	Manhattan	5