Skip to main content
. Author manuscript; available in PMC: 2020 Jan 10.
Published in final edited form as: J Comput Aided Mol Des. 2019 Jan 10;33(1):1–18. doi: 10.1007/s10822-018-0180-4

Figure 8.

Figure 8.

Violin plots of Matthews correlation coefficients between predicted classifications and experimental Kd classifications of active and inactive compounds for submissions that use machine learning and those that don’t in each target dataset: ABL1, JAK2 SC2, JAK2 SC3, p38-a, TIE2, and VEGFR2. Mean, minimum, and maximum Matthews correlation coefficients for each target are shown by whiskers. Null models based on clogP and molecular weight are shown in green and purple, respectively. Null models were not calculated for the ABL1 target, since this subchallenge only contains two ligands.