Table 2. Distributions of performance metrics for models built using each combination of data. Although the mean performance is improved through increasing data integration, the standard deviations indicate strong variability of performance across different selections of training and test data.

Descriptor domains	CCR (mean ± SD)	Sensitivity (mean ± SD)	Selectivity (mean ± SD)
Chemical only	0.78 ± 0.05	0.72 ± 0.11	0.84 ± 0.06
Protein target only	0.67 ± 0.06	0.56 ± 0.12	0.78 ± 0.06
Cytotoxicity only	0.67 ± 0.06	0.40 ± 0.10	0.93 ± 0.03
Chemical and protein target	0.79 ± 0.05	0.74 ± 0.10	0.84 ± 0.06
Chemical and cytotoxicity	0.80 ± 0.06	0.74 ± 0.12	0.86 ± 0.05
Protein target and cytotoxicity	0.73 ± 0.06	0.63 ± 0.13	0.84 ± 0.06
Chemical, protein target and cytotoxicity	0.82 ± 0.05	0.77 ± 0.10	0.86 ± 0.05