ARC (Antigen Receptor Classifier)

Authors: Austin Crinklaw, Swapnil Mahajan

Requirements:

• Linux OS
• HMMER3
• NCBI Blast+
• Python 3+
  • Python packages: Pandas, BioPython

Installation:

We provide a Dockerfile for ease of use.

ARC can also be downloaded through PyPI using the following pip command.

pip install bio-arc

Usage:

Input

• A fasta format file with one or more protein sequences.

>1WBZ_A_alpha I H2-Kb
MVPCTLLLLLAAALAPTQTRAGPHSLRYFVTAVSRPGLGEPRYMEVGYVDDTEFVRFDSDAENPRYEPRARWMEQEGPEYWERETQKAKGNEQSFRVDLRTLLGYYNQSKGGSHTIQVISGCEVGSDGRLLRGYQQYAYDGCDYIALNEDLKTWTAADMAALITKHKWEQAGEAERLRAYLEGTCVEWLRRYLKNGNATLLRTDSPKAHVTHHSRPEDKVTLRCWALGFYPADITLTWQLNGEELIQDMELVETRPAGDGTFQKWASVVVPLGKEQYYTCHVYHQGLPEPLTLRWEPPPSTVSNMATVAVLVVLGAAIVTGAVVAFVMKMRRRNTGGKGGDYALAPGSQTSDLSLPDCKVMVHDPHSLA
>1WBZ_B_b2m I H2-Kb
MARSVTLVFLVLVSLTGLYAIQKTPQIQVYSRHPPENGKPNILNCYVTQFHPPHIEIQMLKNGKKIPKVEMSDMSFSKDWSFYILAHTEFTPTETDTYACRVKHASMAEPKTVYWDRDM

Commands

• Using Fasta file as an input:

python -m ARC classify -i /path/to/input.fasta -o /path/to/output.csv

Output

• Output file has 4 columns in CSV format.
• First column named 'ID' is the description provoded in the fasta for each sequence.
• Second column named 'class' is the assigned molecule class for each sequence.
  • e.g. MHC-I, MHC-II, BCR or TCR.
• The third column named 'chain_type' is the assigned chain type for each sequence.
  • e.g. alpha, beta, heavy, lambda, kappa, scFv, TscFv or construct. These will also be labelled as V for variable domain or C for constant domain.
• The fourth column named 'calc_mhc_allele' is the MHC allele identified using groove domain similarity to MRO alleles.

How it works:

• BCR and TCR chains are identified using HMMs. A given protein sequence is searched against HMMs built using BCR and TCR chain sequences from IMGT. HMMER is used to align an input sequence to the HMMs.
• MHC class I (alpha1-alpha2 domains) and MHC class I alpha and beta chain HMMs are downloaded from Pfam website. An input protein sequence is searched against these HMMs. A HMMER bit score threshold of 25 was used to identify MHC chain sequences. -To identify MHC alleles, MRO repository is downloaded every time the script is run. Groove domains (G-domains) are assigned to new MRO allles and stored in a CSV file. If this file does not exist then G-domains are assigned to all the MRO alleles (which may slow down the script).

References:

Several methods for HMMER result parsing were sourced from ANARCI.

Dunbar J and Deane CM. ANARCI: Antigen receptor numbering and receptor classification. Bioinformatics (2016)