Figure 7.
Distribution of the validity, efficiency, and F1 score values obtained within the 5-fold CV framework for the (a) MNT, (b) DILI, and (c) DICC CP models built on the different descriptor sets after feature selection. The CHEM descriptor set includes the molecular fingerprint and physicochemical descriptors; the BIO descriptor set includes the predicted p-values for a set of biological endpoints (bioactivity descriptor); the CHEMBIO descriptor set includes the previous two descriptor sets. Significant differences in the distribution (p-value <0.05) are denoted by a star.