Assessment of the Generalization Abilities of Machine-Learning Scoring Functions for Structure-Based Virtual Screening

Author: Hui Zhu, Jincai Yang*, and Niu Huang*

Explanation

split_dataset/cluster collected scripts of Pocket Pfam-based clustering (Pfam-cluster) and Protein sequence-based clustering (Seq-cluster). pdbbind_2020_cluster_result.csv contained results of two clustering approaches

split_dataset/3_fold contained the training, validation and testing dataset for generalization ability benchmark in the paper.

models/Descriptor_based_model contained source code of LR::V, LR::VR1, RF-Score, XGB::VR1 and NNScore. Other evaluated models were downloaded from individual paper.

Models	Availability
Pafnucy	http://gitlab.com/cheminfIBB/pafnucy
OnionNet	http://github.com/zhenglz/onionnet/
SG-CNN	https://github.com/llnl/fast
IGN	https://github.com/zjujdj/InteractionGraphNet/tree/master
SIGN	https://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/drug_target_interaction/sign
GraphBAR	https://github.com/jtson82/graphbar

models/shap is the Shapley Additive exPlanations (SHAP) analysis on RF-Score

Citation

Assessment of the Generalization Abilities of Machine-Learning Scoring Functions for Structure-Based Virtual Screening