immunoGAT

immunoGAT is a graph attention transformer model designed to process both structural and sequence information of peptide-MHC complexes for immunogenicity prediction.

Setup:

Clone immunoGAT git repo:

git clone https://github.com/david-zhang03/immunoGAT.git

Environment setup:

If you do not have Anaconda installed, follow these instructions

To verify installation,

conda list

To install environment,

conda env create -f environment.yml

To activate,

conda activate immunoGAT

Usage:

Due to file sizes, first generate graphs following instructions in expand_ablation_graph.ipynb. This allows you to generate graphs from the PDB files found in SingleAllele and Multiallele.

To reproduce the results from the baseline GAT model, follow the instructions in baselineGAT.ipynb.

To reporduce the results from immunoGAT, follow the instructions in immunoGAT.ipynb.