Combined machine learning models, docking analysis, ADMET studies and molecular dynamics simulations for the design of novel FAK inhibitors against glioblastoma

Table 2 Performance of the ML models on the ten-fold cross-validation of test set using LightGBM method

Fingerprint	Input	Test set
Fingerprint	Number	R²	MAE	RMSE
CDK	1024	0.760 ± 0.012	0.417 ± 0.009	0.572 ± 0.014
CDK extended	1024	0.756 ± 0.015	0.421 ± 0.006	0.577 ± 0.018
Substructure	307	0.650 ± 0.009	0.491 ± 0.006	0.691 ± 0.009
Substructure count	307	0.709 ± 0.013	0.462 ± 0.012	0.630 ± 0.014
CDK + CDK extended	2048	0.765 ± 0.010	0.413 ± 0.011	0.565 ± 0.012
Substructure + Substructure count	614	0.710 ± 0.013	0.461 ± 0.012	0.629 ± 0.014
CDK + Substructure count	1331	0.769 ± 0.010	0.398 ± 0.009	0.562 ± 0.012
CDK extended + Substructure count	1331	0.770 ± 0.010	0.412 ± 0.009	0.560 ± 0.012
CDK + Substructure + Substructure count	1638	0.769 ± 0.010	0.398 ± 0.009	0.562 ± 0.012
CDK extended + Substructure + Substructure count	1638	0.770 ± 0.010	0.412 ± 0.009	0.560 ± 0.012
CDK + CDK extended + Substructure	2355	0.765 ± 0.010	0.416 ± 0.010	0.566 ± 0.011
CDK + CDK extended + Substructure count	2355	0.782 ± 0.010	0.398 ± 0.006	0.545 ± 0.013
CDK + CDK extended + Substructure + Substructure count	2662	0.782 ± 0.010	0.398 ± 0.006	0.545 ± 0.013

ISSN: 2661-801X