BioEmu: Conformational Sampling of Annexin A11

Objective

This project performs conformational sampling of the protein backbone for Annexin A11 (both native and mutant variants) using the BioEmu v1.1 generative model. The primary goal is to compare conformational profiles to identify structural changes induced by specific mutations, such as P36R.

Environment & Requirements

• Hardware: NVIDIA GPU (tested on Tesla T4).
• Software:
  • Python 3.10+
  • BioEmu 0.1.6
  • Condacolab (Conda management for Google Colab)
  • MDAnalysis & MDTraj (trajectory processing)
  • AmberTools

Notebook Workflow

1. Setup: BioEmu installation and a compatibility patch for pathlib.py to ensure stability in the Colab environment.
2. Inputs:
   • Annexin A11 FASTA sequence (~505 residues).
   • Mutation string input (e.g., "P36R").
3. Mutation Logic: Automated validation of the original residue at the specified 1-based position before applying the substitution to the sequence string.
4. Execution:
   • Sampling of the native variant.
   • Sampling of the mutant variant.
   • Exporting results to .pdb (structure) and .xtc (trajectory) formats.

Technical Notes & Debugging

• Memory Handling: Due to the length of Annexin A11, the model automatically scales down to 'batch_size = 1' to avoid OOM (Out of Memory) errors.
• IndexError (Zero Size): The logs indicate an IndexError if the trajectory object is empty. This occurs if no samples pass the generative filters or if VRAM limitations prevent coordinate generation.
• Path Compatibility: The script modifies /usr/local/lib/python3.11/pathlib.py to ensure MDAnalysis can correctly parse file paths within the local environment.

Author

João Calisto B.S. Student in Computational Physics - IFSC/USP