pMTnet-Omni-Document 0.0.4

Document for pMTnet Omni

pMTnet Omni: your one-stop TCR-pMHC affinity prediction algorithm :microscope:

Package	Documentation	Code Coverage
pMTnet Omni
pMTnet Omni Document

pMTnet Omni is a deep learning algorithm for affinity prediction based on TCR Va, Vb, CDR3a, CDR3b sequences, peptide sequence, and MHC allele types. The predictions can be made for human and mouse alleles, and for both CD8 T cells/MHC class I and CD4 T cells/MHC class II.

Please refer to our paper for more details: pMTnet Omni paper link here

We host the online tool on DBAI, where you can find all the members of the pMTnet family, including pMTnet V1.

We have also built a detailed online documentation where we guide you step-by-step on how to format your data so it can be accpted by our algorithm.

NOTE: This is the documentation for the data curation supporting tool for pMTnet Omni. Use this BEFORE you upload your dataset to DBAI.

Model Overview

Dependencies

• numpy==1.22.4
• pandas==1.5.2
• tqdm==4.64.1
• torch==1.13.1
• fair-esm==2.0.0

Enviroment Setup

conda env create -f pMTnet_Omni_Document_env.yml

Installation

conda activate pMTnet_Omni_Document
pip install pMTnet_Omni_Document

Quick Start Guide

1. Prepare your dataset so that it looks somewhat like the following: Along with the main program, we also published 5 datasets under the ./validation_data folder. Feel free to use those datasets to check if you TCR namings, Amino Acid sequences, and MHC namings conform with our standard.

NOTE: When both TCR names (resp. MHC) and the TCR sequences (resp. MHC sequences) are provided, we will disregard the sequences. If the names can NOT be found in our reference database, the record WILL be dropped.

NOTE: On the other hand, if the names are NOT provided, we will use the sequences with minimal curation.

2. Say your dataset is under ./df.csv. In your terminal, run

conda activate pMTnet_Omni_Document

python -m pMTnet_Omni_Document --file_path ./df.csv --validation_data_path ./validation_data --output_file_path ./df_result.csv

3. Go to our website and upload your data including the .pickle file.

4. An example output would look like this:

For a more in-depth explanation on input format, check out our online documentation.

CITATION HERE