PreSelAnTest - Epitope prediction, selection and multiple model based antigenicity test

Python3 pipeline to predict t-cell epitopes or parse b-cell bepipred results, filter epitopes and perform multiple-model antigenicity test.

Summary

We have developed a comprehensive pipeline for prediction, simple curation and antigenicity prediction using multiple models. This pipeline contains the following functions: (1) Prediction of t-cell epitopes with netmhcpanII, with desired hla alleles, filtering the input fasta to select aa sequences according to this tool specifications; it also parse the results to organize the input for the other steps of the pipeline. Alternatively, it parses the fasta files aleady predicted by the b-cell bepipred3 predictor for linear epitopes. (2) Curation of these epitopes according to rank percentile of binding affinity (only t-cell), promiscuity of mhc alleles (only t-cell), overlapping with iedb epitopes (only t-cell), overlapping in protegen database bacterial protein sequences, and overlapping of epitopes with human proteins. (3) Antigenicity prediction for these epitopes according to paprec workflow

Requirements:

• Python packages needed:
  • pip3 install numpy
  • pip3 install sklearn=1.2.2
  • pip3 install pandas
  • pip3 install matplotlib
  • pip3 install statistics
  • pip3 install python-Levenshtein
  • pip3 install boruta
  • pip3 install joblib

Usage Instructions

Preparation:

1. git clone https://github.com/YasCoMa/preselantest_pipeline.git
2. cd preselantest_pipeline
3. pip3 install -r requirements.txt
4. Umcompress trained_models.tar.xz (Provided pickle models were generated on python 3.9 with sklearn 1.2.2)
5. The main input file is the configuration in json, there is an example for b-cell (configcarb.json) and for t-cell (config.json). The second parameter of each function file is the step used in (1) and (2). Only in the predictions1.py, there is a third line command argument corresponding to the nemhcpanII tool path, only required if you are using for t-cell epitope prediction

(1) Run Prediction of epitopes or parsing of prediction results:

• Configuration variables:
  
  • folder_in: path to the working directory
  • cell_type: cell type for the epitopes (t-cell or b-cell), value is b or t
  • hlas: it is used only for t-cell, list of mhc alleles
  • bepipred3_output: it is used only for b-cell, bepipred3 fasta output (must be in the working directory)
• Run all:
  • python3 prediction_s1.py config_carb.json 0
• Run only prediction:
  • python3 prediction_s1.py config.json 1 /path/to/nemhcpanII
• Run bepipred3 parsing:
  • python3 prediction_s1.py config_carb.json 1
• Run t-cell results parsing:
  • python3 prediction_s1.py config.json 2 /path/to/nemhcpanII

(2) Run epitope curation:

1. Configuration variables:
   
   • folder_in: path to the working directory
   • cell_type: cell type for the epitopes (t-cell or b-cell), value is b or t
   • thresholdsimiedb: it is used only for t-cell, threshold of similarity between iedb and prediction epitopes
   • threshold_alleles: it is used only for t-cell, threshold of promiscuity of a epitope ith affinity predicted for multiple mhc alleles
   • threshold_rank: it is used only for t-cell, threshold of percentile rank to filter results table from netmhcpanII
2. Run selection:
   • python3 curation_s2.py config_carb.json 1
3. Run vaxijen2 summary output table parsing (only check Summary Mode in web server, copy the results in the webpage and save as vaxijen{epitope|protein}results.txt in {folder_in}/), this step must be made after the selection process:
   • python3 curation_s2.py config_carb.json 2

(3) Run epitope antigenicity prediction:

• Configuration variables:
  
  • folder_in: path to the working directory
  • cell_type: cell type for the epitopes (t-cell or b-cell), value is b or t
  • controlepitope: it is not mandatory, only use if you want to compare the results with vaxijen (if you used step 2 of curations2.py, the value should be vaxijenepitopeprediction.tsv)
  • control_protein: it is not mandatory, only use if you want to compare the results with vaxign-ml
  • For the control files, you may compare with any other tool, you just have to be sure to format according to the model, using the columns:
    • pepid: identifier for epitope/protein
    • sequence: aa sequence
    • value: original score of the tool
    • class: classification in binary number (1 or 0)
    • class_name: antigen or non-antigen
• Run selection:
  • python3 antigenicity_prediction_s3.py config_carb.json

Run all steps

• There is an example of shell file to execute all steps for the config_carb.json file, adapt it according to your configuration file name and the steps you want to run with PreSelAnTest
• chmod a+x run_all.sh
• bash run_all.sh

Reference

Bug Report

Please, use the Issues tab to report any bug.