cghla

A set of tools that are useful for finding the MHC/HLA alleles for a given sample using WGS, WES, or RNAseq data. In particular, these tools are designed for the identification of MHC alleles to use in identification of novel peptides that can be presented by MHC Class I proteins for finding presented neoepitopes in cancer. As such, the specificity of the results are designed only for this purpose.

These tools implement a particular workflow that assumes you have the following:

bwa
samtools
ngsutilsj

Additionally, the tools are built to work with data from the IPD-IMGT/HLA database which is available from: https://hla.alleles.org

Finally, final prediction can use MHC peptide binding affinities from mhcFlurry which are available here: https://openvax.github.io/mhcflurry-motifs/mhcflurry.allele_sequences.csv

The basic workflow can be described as:

WGS

1. Find the location of HLA alleles in the reference genome.
2. Extract flanking sequence from the reference genome, and adding this sequence to the allele sequences
3. Extraction of mapped reads to the HLA locus in an already aligned BAM file
4. Re-alignment of these reads (optionally with unmapped reads as well) to the now extended flanking HLA allele sequences
5. Score HLA allele pairs (maternal/paternal) to find the most likely genotypes for the individual.
6. Predict the most likely genotypes

RNAseq

The RNAseq/WES workflow is similar, but instead of using the *_gen.fasta versions of the IPD-IMGTR/HLA sequences you can use the *_nuc.fasta versions. When using these coding mRNA sequence, you can skip steps 1 and 2. And instead of extracting a sub-set of reads, you can align the raw FASTQ sequence. The scoring and prediction steps remain the same.