Programme

P.60 -- Poster, VCTP-49

Date: Friday, 2 August 2024

Time: 08:30 - 10:00

Evaluation of Pafnucy deep neural network model for ligand based drug virtual screening

Nguyen Tien Cuong, Khuat Dang Son, Nguyen Nhat Tung, Nguyen Viet Anh, and Nguyen The Toan

Faculty of Physics, VNU University of Science, Hanoi

Pafnucy is a 3D convolutional neural network that predicts binding affinity for protein-ligand complexes developed by Marta et. al. The original version was trained on the PDBbind database in 2016 and tested on the CASF "scoring power" benchmark. In this study, we retrain the Pafnucy model with more updated data. The Pafnucy models were then used to predict protein-ligand affinity on SARS-COV-2 and mu-opioid receptors. The calculated results show that Pafnucy models can predict the protein-ligand binding affinity very fast with a good accuracy. Our retrained Pafnucy model gives results with higher accuracy than the original version. Pafnucy has great potential when combined with traditional computational tools such as molecular dynamics and docking to speed up the virtual drug screening process.

Presenter: Nguyen Tien Cuong