mhdx_analysis

This repository contains scripts and notebooks to analyze results from mhdx_pipeline and hdxrate_pipeline associated with the Ferrari 2025 paper.

Table of Contents

1. Compute Cooperativities
2. AF2 Prediction
3. Rosetta Relaxation
4. Hydrogen Bond Extraction
5. Dependencies
6. How to Cite

---

Compute Cooperativities

The code to compute cooperativities (normalized cooperativity and family-normalized cooperativity) from scratch can be found in:

Path: notebooks/DeriveCooperativities.ipynb

Requirements:

• Processed Data: Output from hdxrate_pipeline ({hdxrate_output}/consolidated_results/deduplicated.json).
• Structural Data: Dataframe with protein names and the number of expected protected amides based on their 3D structure.

Structural Modeling Details:

• Structures from Ferrari 2025 were generated using AlphaFold2 via ColabFold, followed by RosettaRelax applied to the top-scoring model.
• Hydrogen bond information was extracted using a custom PyRosetta script.
• Protein family info was added by populating a PF column to the final hb.json dataframe

Scripts for all these steps are provided. Ensure ColabFold and Rosetta are installed. Scripts include the exact parameters used in our study.

Script to populate cooperativity metrics from precomputed average trends found in Ferrari 2025

mhdx_pipeline/scripts/cooperativity/resources/20241030_param_table.json and mhdx_pipeline/scripts/cooperativity/resources/240917_cooperativity_std_mean_dict.json are provided

python mhdx_pipeline/scripts/cooperativity/compute_cooperativity.py --hx deduplicated.json --hb hb.json --param_table  mhdx_pipeline/scripts/cooperativity/resources/20241030_param_table.json --cooperativity_dict  mhdx_pipeline/scripts/cooperativity/resources/240917_cooperativity_std_mean_dict.json --output results/df_HX_cooperativities.json

---

AF2 Prediction

To perform structure prediction using AlphaFold2:

bash mhdx_analysis/scripts/feature_extraction/run0_structure_prediction.sh {folder-to-fasta-or-a3m-files}

---

Rosetta Relaxation

To relax the AlphaFold2 predicted structures using Rosetta:

bash mhdx_analysis/scripts/feature_extraction/run1_rosetta.sh {af_prediction/pdbs/}

---

Hydrogen Bond Extraction

Extract hydrogen bond information from the relaxed structures.

Run for each structure:

python mhdx_analysis/scripts/feature_extraction/pdb2hbond.py --input "$rosetta_relax" --output "$hbonds_output"

After extraction, concatenate all H-bond JSON files and populate a column PF with the corresponding protein family.

---

Dependencies

Ensure the following tools and libraries are installed:

• ColabFold (version 1.5.2): https://github.com/sokrypton/ColabFold
• Rosetta (version 2019.11) : https://rosettacommons.org/software/
• PyRosetta (version 4 2022.33) : https://www.pyrosetta.org/

---

How to Cite

If you use this repository, please cite the following paper:

Ferrari, A. et al. (2025). Title of the Paper. Journal Name. DOI: XXXXXX

---

For any questions or issues, please contact [email protected].