Predict mutated T-cell epitopes from sequencing data
Python Shell
Latest commit b7a4f2a Jan 17, 2017 @julia326 julia326 committed on GitHub Merge pull request #64 from hammerlab/fix-test-after-netmhcpan-update
Fixing test after netMHCpan 2.8 -> 3.0 update.

Build Status Coverage Status DOI


Predict mutation-derived cancer T-cell epitopes from (1) somatic variants (2) tumor RNA expression data, and (3) patient HLA type.


./topiary \
  --vcf somatic.vcf \
  --mhc-predictor netmhcpan \
  --mhc-alleles HLA-A*02:01,HLA-B*07:02 \
  --ic50-cutoff 500 \
  --percentile-cutoff 2.0 \
  --mhc-epitope-lengths 8-11 \
  --rna-gene-fpkm-tracking-file genes.fpkm_tracking \
  --rna-min-gene-expression 4.0 \
  --rna-transcript-fpkm-tracking-file isoforms.fpkm_tracking \
  --rna-min-transcript-expression 1.5 \
  --output-csv epitopes.csv \
  --output-html epitopes.html


You can install Topiary and all of the libraries it depends on by running:

pip install topiary

You'll need to download the reference genome sequences and annotations for a recent Ensembl release (e.g. 81) by running:

pyensembl install --release 81 --species human

If you want to work with variants which were aligned against the older reference GRCh37, you will need to also download its annotation data, which is contained in Ensembl release 75:

pyensembl install --release 75 --species human

Commandline Arguments

Genomic Variants

Specify some variants by giving at least one of the following options. They can be used in combination and repeated.

  • --vcf VCF_FILENAME: Load a VCF file
  • --maf MAF_FILENAME: Load a TCGA MAF file
  • --variant CHR POS REF ALT : Specify an individual variant (requires --ensembl-version)

Output Format

  • --output-csv OUTPUT_CSV_FILENAME: Path to an output CSV file
  • --output-html OUTPUT_HTML_FILENAME: Path to an output HTML file

RNA Expression Filtering

Optional flags to use Cufflinks expression estimates for dropping epitopes arising from genes or transcripts that are not highly expressed.

  • --rna-gene-fpkm-tracking-file RNA_GENE_FPKM_TRACKING_FILE: Cufflinks FPKM tracking file containing gene expression estimates.
  • --rna-min-gene-expression RNA_MIN_GENE_EXPRESSION: Minimum FPKM for genes
  • --rna-transcript-fpkm-tracking-file RNA_TRANSCRIPT_FPKM_TRACKING_FILE: Cufflinks FPKM tracking file containing transcript expression estimates.
  • --rna-min-transcript-expression RNA_MIN_TRANSCRIPT_EXPRESSION: Minimum FPKM for transcripts
  • --rna-transcript-fpkm-gtf-file RNA_TRANSCRIPT_FPKM_GTF_FILE: StringTie GTF file file containing transcript expression estimates.

Choose an MHC Binding Predictor

You must choose an MHC binding predictor using one of the following values for the --mhc-predictor flag:

  • netmhc: Local NetMHC predictor (Topiary will attempt to automatically detect whether NetMHC 3.x or 4.0 is available)
  • netmhcpan: Local NetMHCpan predictor
  • netmhciipan: Local NetMHCIIpan predictor
  • netmhccons: Local NetMHCcons
  • random: Random IC50 values
  • smm: Local SMM predictor
  • smm-pmbec: Local SMM-PMBEC predictor
  • netmhcpan-iedb: Use NetMHCpan via the IEDB web API
  • netmhccons-iedb: Use NetMHCcons via the IEDB web API
  • smm-iedb: Use SMM via the IEDB web API
  • smm-pmbec-iedb: Use SMM-PMBEC via the IEDB web API

MHC Alleles

You must specify the alleles to perform binding prediction for using one of the following flags:

  • --mhc-alleles-file MHC_ALLELES_FILE: Text file containing one allele name per line
  • --mhc-alleles MHC_ALLELES: Comma separated list of allele names, e.g. "HLA-A02:01,HLA-B07:02"

Peptide Length

  • --mhc-epitope-lengths MHC_EPITOPE_LENGTHS: comma separated list of integers specifying which peptide lengths to use for MHC binding prediction

Binding Prediction Filtering

  • --only-novel-epitopes: Topiary will normally keep all predicted epitopes, even those which occur in a given self-ligandome or don't overlap a mutated region of a protein. Use this flag to drop any epitopes which don't contain mutations or that occur elsewhere in the self-ligandome.
  • --ic50-cutoff IC50_CUTOFF: Drop peptides with predicted IC50 nM greater than this value (typical value is 500.0)
  • --percentile-cutoff PERCENTILE_CUTOFF: Drop peptides with percentile rank of their predicted IC50 (among predictions for a particular allele) fall below this threshold (lower values are stricter filters, typical value is 2.0)


  • --padding-around-mutation PADDING_AROUND_MUTATION: Include more unmutated residues around the mutation (useful when not using --only-novel-epitopes)
  • --self-filter-directory SELF_FILTER_DIRECTORY: Directory of files named by MHC allele containing a self peptide ligandome (peptides which should be excluded from results)
  • --skip-variant-errors: If a particular mutation causes an exception to be raised during annotation, you can skip it using this flag.