"Advancing Protein Ensemble Predictions Across the Order-Disorder Continuum" Repository Overview

This repository provides all data, models, and results associated with the PepTron study on protein structure prediction for ordered and disordered proteins described in https://www.biorxiv.org/content/10.1101/2025.10.18.680935. 

The code for PeptoneBench can be found at https://github.com/PeptoneLtd/PeptoneBench.
The code for the PepTron model can be found at https://github.com/PeptoneLtd/PepTron.
The code for the IDP-o model can be found at https://github.com/PeptoneLtd/IDP-o.

Repository Structure

• PepTron: Contains the checkpoint of the PepTron model
• PeptoneDBs: Experimental data curated from the literature (PeptoneDB-CS, PeptoneDB-Integrative,PeptoneDB-SAXS), covering both ordered and disordered proteins. Includes raw datasets and sequence information.
• Predictions: Protein structure predictions for entries in PeptoneBench, generated using PepTron and other state-of-the-art models (BioEmu, Alphafold2, Boltz, etc.). Each entry includes 3D conformations in PDB and XTC formats, and back-calculated SAXS and chemical shift (CS) data from predicted structures.
• IDRome-o: Collection of 3D structures from IDRome, generated with the IDP-o model. These structures were used for fine-tuning PepTron.