QuickMHC

Wrapper around NetMHCpan3.0 and a C++ implementation of our peptide similarity to self determination code for running on *nix systems. Results are stored in a PostgreSQL database for future lookups. The $USER is assumed to have read/write access to this database, named binding_affinity.

Binding affinity predictions are stored as such (example from the A0201 table)

peptide     peptide_score_log50k   affinity   percentile_rank
ELVISLIVE   2.48e+04               6.48e-02   29
SYFPEITHI   1.55e-01               9.32e+03   13

STS predictions are stored as such (example from the STS_A1101_1.9 table)

peptide     different_from_self
SYFPEITHI   TRUE
ELVISLIVE   FALSE

Installation

• Install PostgreSQL on your system, initialize a database named "binding_affinity" and couple a database user to your Linux user account

I.e. start psql from the shell and type:

CREATE DATABASE binding_affinity;
CREATE USER my_user_name;

• Clone this repo to a local directory and compile and install this package, probably most easily done by running build_package.sh in the root directory

• Configure quickMHC to know what user and database to use. You can do this by defining any config file (found using the regex: .*quickMHC.*\\.yaml) in the following directories (ordered in descending priority): ~/.config, the current working directory, the location of quickMHC's installation.

Example contents of this file:

db_user: m.slagter # Or whatever db_user you created in the first step
db_name: binding_affinity

Usage

• Load quickMHC in R session
• Create quickMHC object and query peptides of interest

## For binding affinity predictions
library(quickMHC)
BA_predictor <- BindingPredictor$new(hla_allele = 'A0201')
## Peptide will be computed the first time around, will be looked up from database in 
## subsequent queries
BA_predictor$query(c('SYFPEITHI'))
# peptide     peptide_score_log50k   affinity   percentile_rank
# SYFPEITHI   1.55e-01               9.32e+03   13

## For STS predictions
STS_predictor <- STSPredictor$new(hla_allele = 'A0201')
BA_predictor$query(c('SYFHPETHI'))
# peptide     different_from_self
# ELVISLIVE   FALSE

Todo

• Make location of NetMHCpan configurable
• Add functionality to create self lists for STS computation. Low priority, since STS is not contributing much predictive power.
• Allow for user defined 'predictor' wrapper functions
• A Singulariy/Docker container with both this package and the database pre-configured