Neoepitope novelty

These tools provide annotation of neoepitope predictions to allow for insights into the level of novelty of a neoepitope. We produce metrics regarding novelty of MHC binding and of peptide sequence relative to human, bacterial, and viral peptides. Our manuscript in BMC Cancer demonstrates the utility of these metrics for predicting immunogenicity of neoepitopes.

Steps for use:

1) Download required software:

• BLAST+ (v2.4.0+)
• NetMHCpan (v2.8)
• Python (v2.7+)
• Python libraries: BioPython (v1.70)

You will also need a neoepitope prediction software! We used and recommend pVAC-Seq (v4.0.8) for this analysis, as this code best supports that functionality. However, you're welcome to use any software you like, it will just require a bit more processing on your end!

2) Obtain and process required files:

a) Download and unzip Refseq's nonredundant bacterial peptide fasta set: ftp://ftp.ncbi.nlm.nih.gov/refseq/release/bacteria/bacteria.nonredundant_protein.*.protein.faa.gz

Note: our code works for release #83

Concatenate all the bacterial fasta files into one combined, bacterial peptide fasta file

b) Run prepare_dbs.py from within the neoepitope_novelty/ directory to produce a blast protein database from the combined bacterial fasta file:

python prepare_dbs.py -m <makeblastdb path> -d blast_dbs/ -b <bacterial fasta>

• -p path to makeblastdb executable
• -d path to directory in which write database
• -b path to bacterial peptide fasta (produced as in step 2a)

c) Download the pickled dictionary of bacterial peptides, bacterialDict.pickle, available on figshare: https://figshare.com/articles/Peptide_ID_to_species_of_origin_dictionary/5345323

Move the dictionary file into the neoepitope_novelty/dictionaries/ directory

3) For each sample and allele, run pVAC-Seq (v4.0.8) according to their recommendations:

http://pvac-seq.readthedocs.io/en/v4.0.8/index.html

You may choose any epitope length of your choice, but otherwise use default parameters.

You may opt to use different parameters or a different tool, but you will then need to parse your results independently. (See step 4 for instructions.)

4) For each run of pVAC-Seq, parse the results to produce a reduced results file:

If pVAC-Seq was run according to our recommendations above, you can use parse_pvac.py to produce the formatted file:

python parse_pvac.py -i <input file> -o <output file> -s <sample name>

• -i input file, the path to the results from running pVAC-Seq with the recommended settings
• -o output file, the desired path to write results to
• -s sample name to distinguish run

If you chose to use different parameters for running pVAC-Seq, or used a different program, that's fine too! You'll just need to parse the results independently. Produce a tab-delimited file from your results with no header and with the following columns in this order:

<allele> <neoepitope sequence> <neoepitope binding affinity> <normal epitope sequence> <normal epitope binding affinity> <transcript> <gene>

5) For each sample, run annotate_neoepitopes.py:

python annotate_neoepitopes.py -i <input file> -o <output directory> -s <sample name> -a <allele> -p <blastp path> -n <netMHCpan path>

• -i input file to program, the path to the result of parsing a pVAC-Seq results file with parse_pvac.py
• -o path to output directory to which output files will be written
• -s sample name to distinguish run
• -a HLA allele to use for analysis (format example: HLA-A02:01)
• -p path to blastp+ executable
• -n path to netMHCpan executable

Results will be in your chosen output directory under the name [SAMPLE NAME].epitopes.annotated.tsv

6) If you use these tools in analyses for your publication, please cite our manuscript:

Wood MA, Paralkar M, Paralkar MP, Nguyen A, Struck AJ, Ellrott K, Margolin A, Nellore A, Thompson RF. (2018) Population-level distribution and putative immunogenicity of cancer neoepitopes. BMC Cancer, 18, 414.