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

Testing Installation:

A quick check for proper dependencies and successful installation can be performed by navigating to your pip package install directory (which can be located by executing pip show bio-arc) and running the following command:

python3 -m arc_test

Passing all unit-tests means that your system is configured properly and ready to classify some protein sequences.

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.

ID	class	chain_type	calc_mhc_allele
1WBY_A_alpha I H2-Db	MHC-I	alpha V
1WBY_B_b2m I H2-Db
1HQR_A_alpha II HLA-DRA01:01/DRB501:01	MHC-II	alpha C	HLA-DRA*01:01
1HQR_B_beta II HLA-DRA01:01/DRB501:01	MHC-II	beta C	HLA-DRB5*01:01
2CMR_H_heavy	BCR	heavy V
2CMR_L_light	BCR	kappa C
4RFO_L_light	BCR	lambda V
3UZE_A_heavy	BCR	scFv
1FYT_D_alpha	TCR	alpha V
1FYT_E_beta	TCR	beta C
3TF7_C_alpha	TCR	TscFv

Building and using the Singularity image

Building the singularity image requires root-level access and should thus be built on a machine where you have such access. Once it's built, it can be run by any non-root user and can be transferred to other machines. To build:

singularity build arc.sif Singularity

The input and output directories need to be made available to the running container. If these directories are not within your home directory or the directory from which you will be running the container ($PWD), you will need to bind mount these directories in your call to the 'singularity run' command. Otherwise, usage is identical to the non-containerized version:

singularity run \
--writable-tmpfs \
--bind /path/to/host_dir:/host \
arc.sif python3 ARC -m classify -i /host/input_file.fasta -o /host/output_file.tsv

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, groove domains (G-domains) are assigned based on the MRO repository.
IgNAR sequences are identified through querying against a custom blast database.

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)

danielmarrama/ARC