Back

DeepDOX1: A Dual-Drive Framework Integrating Deep Learning and First-Principles Physics for Drug-Protein Affinity Prediction

Liu, Z.; Sun, H.; Wang, Y.; Ren, Y.; Rao, L.; Huang, Z.; Cao, H.; Hu, X.; Zhu, X.; Li, M.; Wan, J.

2025-12-15 bioinformatics
10.64898/2025.12.12.693818 bioRxiv
Show abstract

In this work, we present DeepDOX1, a dual-drive drug-protein affinity (DPA) prediction tool features the tight integration of a concise AI architecture and a quantum mechanics based representation. The first-principle physics generated features incorporating the interactions between the drug and the protein pocket allows a relatively simple CNN model trained on a relatively small training set to exhibit exceptional generalization capabilities across extensive testing. Notably, DeepDOX1 outperforms popular AI models in the tests simulating real-world drug design scenario and a highly challenging test set featuring covalent ligands, halogenated ligands and metalloproteins. In addition, we designed a series of novel covalent inhibitors targeting the diabetes target hu-FBPase using DeepDOX1. Subsequent experimental validation including enzyme-level bioactivity assays and crystal structure determination confirmed the DeepDOX1s effectiveness in real-world drug design applications. It is conceivable that the combination of AI and first-principle QM might be one of the next breakthrough points of DPA prediction.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.