GraphPrint: Extracting Features from 3D Protein Structure for Drug Target Affinity Prediction

TL;DR

GraphPrint introduces a novel framework that leverages 3D protein structure features represented as graphs, combined with drug features, to enhance drug target affinity prediction accuracy.

Contribution

This work is the first to incorporate 3D protein structure graph features into drug affinity prediction models, improving prediction performance over traditional methods.

Findings

Achieved a mean square error of 0.1378 on KIBA dataset.

Attained a concordance index of 0.8929, outperforming traditional feature-based models.

3D structure features provide complementary information to traditional features.

Abstract

Accurate drug target affinity prediction can improve drug candidate selection, accelerate the drug discovery process, and reduce drug production costs. Previous work focused on traditional fingerprints or used features extracted based on the amino acid sequence in the protein, ignoring its 3D structure which affects its binding affinity. In this work, we propose GraphPrint: a framework for incorporating 3D protein structure features for drug target affinity prediction. We generate graph representations for protein 3D structures using amino acid residue location coordinates and combine them with drug graph representation and traditional features to jointly learn drug target affinity. Our model achieves a mean square error of 0.1378 and a concordance index of 0.8929 on the KIBA dataset and improves over using traditional protein features alone. Our ablation study shows that the 3D protein structure-based features provide information complementary to traditional features.