Skip to main content

Table 2 The statistical performance of SMPLIP-DNN models on PDBbind (Release 2015) according to different features compositions

From: SMPLIP-Score: predicting ligand binding affinity from simple and interpretable on-the-fly interaction fingerprint pattern descriptors

Features Train Valid Test
LOSS RMSE PCC LOSS RMSE PCC LOSS RMSE PCC
IFP 0.563 0.871 0.899 1.019 1.483 0.678 1.032 1.538 0.726
IFP + Int-Dist 0.472 0.990 0.873 0.775 1.468 0.687 0.805 1.582 0.713
IFP + Frag 0.209 0.595 0.956 0.631 1.402 0.699 0.646 1.530 0.733
IFP + Int-Dist + Frag 0.212 0.403 0.979 0.834 1.400 0.733 0.923 1.559 0.714
  1. The Refined set (n = 3481) used for training and validation, and core set (n = 180) as a test data. The boldface represents the model with better statistics from different features combination