February 5, 2015 Scott Brown

Citation: Brown et al. Genome Medicine 2015, 7:125 doi: 10.1186/s13073-015-0248-x http://www.genomemedicine.com/content/7/1/125

NOTE: See below for information on simulating TCR transcripts.

The MiTCR tool does not use read pairing information, so all sequence data is first pooled into a single fastq file. During creation of this pooled read file, reads are checked that they only contain ACTGN bases, and are at least 40 bases in length.

Space requirements: This pipeline uses 10GB RAM. This pipeline will create a copy of the sequence data, so will double the space requirement.

Pipeline:

All fastq files for a single sample, unzipped and uncompressed, must reside in a single directory.

~$ python2 combineAndCleanFastq.py /path/to/fastq/files/ /working/directory/

This will output /working/directory/reads.fq and /working/directory/numReads.txt

• reads.fq: Cleaned, combined RNA-seq file for input into MiTCR.
• numReads.txt: The number of RNA-seq reads to be used for MiTCR.

To run MiTCR on RNA-seq data, the correct parameter set must be selected. This is based on TCR chain and read length. Parameter files are in the format: rnaseq[READ_LEN]bp[TCR_CHAIN].pset

Choose READ_LEN (50, 76, 101) that is closest to actual read length Choose TCR_CHAIN (TRA, TRB) that matches chain you are interested in.

NOTE: mitcr.jar obtained from github.com/milaboratory/mitcr/releases

~$ java -Xmx10g -jar mitcr.jar -pset rnaseq50bpTRA.pset /working/directory/reads.fq /working/directory/outputTRA50bp.txt

And the other chain...

~$ java -Xmx10g -jar mitcr.jar -pset rnaseq50bpTRB.pset /working/directory/reads.fq /working/directory/outputTRB50bp.txt

Output files contain all CDR3 sequences that were extracted.

###Simulating TCR transcripts.

File: simulate_TCR_transcripts.py

Requires Python 2.

Usage: python2 simulate_TCR_transcripts.py /path/to/references/ outputFile.fq numberOfAlphaBetaPairs

Directory containing references expects a separate file for each gene, ex TRAV.fa, TRAJ.fa, TRAC.fa, etc. Different alleles for a gene are contained within the file, and have the entire sequence on a single line.

Quality scores for fastq are set uniformly high ("J").

TCR genes are selected randomly from a uniform distribution. Junctional nucleotide additions and subtractions are selected based on observed frequencies.