ProtXAI: Explainable AI Analysis of Protein Dynamics

• Manuscript: ProtXAI: Explainable AI Reveals Structural Determinants of Protein Dynamics
• Zenodo repository: https://zenodo.org/records/21872520
• Loschmidt Laboratories: https://loschmidt.chemi.muni.cz/

The repository contains the code and implementation for three case studies using the ProtXAI pipeline for the analysis of molecular dynamics (MD) data. The framework combines machine learning and explainable AI (XAI) to identify structural determinants of protein dynamics.

Requirements

The analyses are implemented in Python and use libraries including:

• NumPy
• pandas
• SciPy
• scikit-learn
• TensorFlow / Keras
• PyEMMA
• MDTraj
• BioPython
• Matplotlib
• iNNvestigate for explainable AI and Layer-wise Relevance Propagation (LRP)
• Seaborn

Some preprocessing steps additionally require cpptraj.

For information about iNNvestigate and installation, see the iNNvestigate repository.

Case studies

ApoE4

CNN-based classification of ApoE4 and ApoE4 + 3-SPA using residue–residue distance-difference features, followed by LRP analysis.

Main file: Clean code_ApoE4.py

See ApoE4/README.md for details.

SAK

CNN-based classification of four SAK variants using residue–residue distance-difference features, followed by LRP analysis.

Main file: SAK_clean code.py

See SAK/README.md for details.

Luciferase

Convolutional autoencoder for next-snapshot prediction using Cα coordinates, followed by LRP analysis of the learned protein dynamics.

Main file: Clean_anc.py

See LUC/README.md for details.

Contact

Faraneh Haddadi Loschmidt Laboratories, Masaryk University Email: faranehhaddadi@gmail.com