Trained Model Weights for OmniBind: Pan–Pharmacological Drug–Target Interaction Prediction with 3D–Informed Protein Encoding at Scale

Read full text Summarise with VIRAT
Authors: Akinori Kawaharada, Takumi Ito, Hideyuki Shimizu
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

### OmniBind: Pre-trained model checkpoints for compound-protein interaction prediction This repository contains the pre-trained model weights for **OmniBind**, a unified framework for pan-pharmacological prediction of drug-target interactions (DTIs) using 3D-structure-informed adaptive gated fusion. OmniBind simultaneously predicts four key binding affinity metrics (**Ki, Kd, IC50, and EC50**) from:- Compound SMILES- Protein amino acid sequences- Foldseek 3Di structural alphabet sequences #### Included CheckpointsThe provided weights are organized to ensure both scientific reproducibility and practical utility: - **benchmark/seed[42, 123, 369, 777, 2024].pth**: These five checkpoints correspond to the performance evaluation results (mean ± std) reported in the associated paper. Each is the checkpoint with the lowest validation RMSE for its respective data split; test sets were not used for model selection.- **application.pth**: The model used for the application analyses in the paper (drug repositioning, attention analysis and off-target screening). It is the checkpoint of the primary data split (seed 369) that was used throughout model development, and contains the same weights as `benchmark/seed369.pth`. #### UsageTo use these weights, download the files and place them into the `checkpoints/` directory of the OmniBind repository. For installation, environment setup, and detailed usage instructions, please visit our GitHub repository:https://github.com/Shimizu-team/OmniBind #### Associated Publication**"Pan–Pharmacological Drug–Target Interaction Prediction with 3D–Informed Protein Encoding at Scale"**

Fields of Study & Subjects

Computational Drug Discovery Methods Machine Learning in Bioinformatics Protein Interaction Studies and Fluorescence Analysis

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
30 March 2026
Language
EN
Document type
Dataset