neoantigen-dev

MHC Class I neoantigen prediction pipeline from IM5/IM6/WES/WGS. Takes Normal .bam and somatic .maf and generates neoantigen predictions for HLA-A/B/C.

The pipeline has four main steps:

1. Genotype HLA. genotyping performed using POLYSOLVER.
2. Construct mutated peptides. For non-synonymous mutations, generates mutated peptide sequences based on HGVSc. NOTE: .maf file should be VEP annotated using cmo_maf2maf --version 1.6.14 --vep-release 88 using this EXACT VERSION. TODO: Generate mutated sequences for fusions.
3. Run NetMHCpan-4.0 and NetMHC-4.0. using default parameters for each algorithm.
4. Post-processing. compiles predictions from both algorithms and finds strongest binder for each non-synonymous mutation. Also, each predicted neopeptide is searched against the entire reference peptidome to make sure it is a true neopeptide. is_in_wt_peptidome column reflects that. TODO: Incorporate neoantigen quality from Lukzsa et al., Nature 2017

Install

Clone the repo and install any necessary Python libraries from requirements.txt. This repo is currently compatible with Python 2 and 3. To install:

git clone https://github.com/taylor-lab/neoantigen-dev.git
cd neoantigen-dev
pip install -r requirements.txt

Usage

NOTE: For POLYSOLVER step, the pipeline requires 8 cores.

# Neoantigen prediction pipeline. Four main steps:
		(1) Genotype HLA using POLYSOLVER,
		(2) Constructed mutated peptide sequences from HGVSp/HGVSc
		(3) Run NetMHC-4.0 + NetMHCpan-4.0
		(4) Gather results and generate:
				- <sample_id>.neoantigens.maf: original maf with neoantigen prediction columns added
				- <sample_id>.all_neoantigen_predictions.txt: all the predictions made for all peptides by both the algorithms

optional arguments:
  -h, --help            show this help message and exit

Required arguments:
  --config_file CONFIG_FILE
                        See: neoantigen-luna.config in the repo
  --sample_id SAMPLE_ID
                        sample_id used to limit neoantigen prediction to
                        identify mutations associated with the patient in the
                        MAF (column 16).
  --output_dir OUTPUT_DIR
                        output directory
  --maf_file MAF_FILE   expects a CMO maf file (post vcf2maf.pl)
  --normal_bam NORMAL_BAM
                        full path to normal bam file. Either --normal_bam or
                        --hla_file are required.

Optional arguments:
  --hla_file HLA_FILE   POLYSOLVER output file (winners.hla.txt) for the
                        sample. If not provided,POLYSOLVER is run. Either
                        --normal_bam or --hla_file are required.
  --peptide_lengths PEPTIDE_LENGTHS
                        comma-separated numbers indicating the lengths of
                        peptides to generate. Default: 9,10
  --keep_tmp_files      keeps POLYSOLVER's temporary files. for debugging
                        purposes. Note: TMP files can be more than 5GB.
                        Default: true
  --force_rerun_polysolver
                        ignores any existing polysolver output and re-runs it.
                        Default: false
  --force_rerun_netmhc  ignores any existing netMHCpan output and re-runs it.
                        Default: false

Output

HLA genotypes

<output_dir>/polysolver/winners.hla.txt

Neoantigen binding affinties annotated MAF

<sample_id>.neoantigens.maf. (peptide with the highest binding affinity is incorporated into the original .maf for each non-syn mutation)
<sample_id>.all_neoantigen_predictions.txt: all the predictions made for all peptides by both the algorithms

The following columns are appended to the input .maf.

Column Name	Description
neo_maf_identifier_key	a unique key that can be used to find other peptides predicted for the same mutation (in .all_neoantigen_predictions.txt)
neo_best_icore_peptide	neopeptide sequence for the strongest binder
neo_best_rank	binding rank for the strongest binder
neo_best_binding_affinity	binding affinity for the strongest binder
neo_best_binder_classification	binding classification for the strongest binder (Non Binder, Strong Binder, Weak Binder)
neo_best_is_in_wt_peptidome	TRUE/FALSE indicating whether the strongest binder peptide is in the reference peptidome
neo_best_algorithm	algorithm predicting the strongest binder
neo_best_hla_allele	hla allele for the strongest binder
neo_n_peptides_evaluated	total # of all peptides evaluated (unique icore peptides)
neo_n_strong_binders	total # of strong binders
neo_n_weak_binders	total # of weak binders

The column description for .all_neoantigen_predictions.txt can be found in: http://www.cbs.dtu.dk/services/NetMHC/output.php. Additional columns are:

The following columns are appended to the input .maf.

Column Name	Description
binder_class	Non Binder, Strong Binder (rank < 0.5 or affinity < 50), Weak Binder (rank < 2 or affinity < 500)
best_binder_for_icore_group	TRUE/FALSE indicating if the binding prediction is the strongest among all the HLA-alleles/algorithms for the given icore peptide.
is_in_wt_peptidome	if the peptide is present in any other protein in the entire peptidome
neo_maf_identifier_key	a unique key that can be used to find other peptides predicted for the same mutation (in .neoantigens.maf)

Example

python neoantigen.py --config_file neoantigen-luna.config \
                     --sample_id <sample_id> \
                     --normal_bam <normal.bam> \
                     --output_dir <output_dir> \
                     --maf_file <cmo_vep_annotated_maf_file>