Vaxrank '--output-neoepitope-report' fails on StopLoss variants

[JPFinnigan]

Vaxrank Neoepitope Report fails during .xlsx creation with an error stating that the WT MHC binding affinity prediction must be a number. With the same input data sources the vaccine peptide report is produced. When this, and other StopLoss variants are removed from the corresponding .vcf the task runs to completion and produces the expected .xlsx.

Cross-referencing the vaccine peptide report, which handles StopLoss variants I believe this arises for mutant peptides downstream of the reference stop codon. For these peptides it seems Vaxrank doesn't generate a corresponding WT peptide, and states this as text. During xlsx generation it seems you try to append ~ "nM" to the predicted IC50, but this fails w/ non-numbers.

The following variants from the B16F10 alignments toGRCm38/mm10 should reproduce the error:

• 18:36683093A>C
• 5:122320874A>G
• 8:91220179A>T