Prakriti's Hacking of

Mitigating Inequity in MHC Binding Predictions for T Cell Epitope Discovery

Motivation: Computational tools that predict peptide binding by major histocompatibility complex (MHC) proteins play an essential role in current approaches to harness adaptive immunity to fight viral pathogens and cancers. However, there are >22,000 known class-I MHC allelic variants, and it is unknown how well binding preferences are predicted for most alleles. We introduce a machine learning framework that enables state-of-the-art MHC binding prediction along with per-allele estimates of predictive performance.

Results: We demonstrate stark disparities in how much binding data are associated with HLA alleles of individuals across racial and ethnic groups. Pan-MHC modeling mitigates some of these disparities when predicting MHC-peptide binding, and we devise a strategy to begin to address remaining inequities by leveraging our per-allele predictions of performance. The approaches introduced here further the development of equitable MHC binding models, which are necessary to understand adaptive immune response and to design effective personalized immunotherapies in genetically diverse individuals.

MHCGlobe & MHCPerf Installation

MHCGlobe and MHCPerf are both easily accessible for model inference and re-training.

1. Download the mhcglobe git repository containing the code:

   git clone https://github.com/ejglynn/mhcglobe.git

2. Update the mhcglobe_dir variable in src/paths.py with the full path to your mhcglobe folder.

3. From the mhcglobe folder create and activate a Python3 virtual environment with the following commands:

   python3 -m pip install --user --upgrade pip

   python3 -m pip install --user virtualenv

   python3 -m venv env

   source env/bin/activate

4. Install prerequisites in the virtual environment:

   pip3 install jupyter pandas scipy sklearn tensorflow tqdm

5. From the mhcglobe folder, start jupyter:

   jupyter notebook

On your browser, click on the MHCGlobe_User_Notebook.ipynb to open and interact with the notebook.

To speed things up, output files have already been provided in the output folder. If you want to recompute these files, simply delete or rename the output folder.