Add support for Combinatorial library and Sturniolo (tepitope) Class II prediction algorithms

[jhundal]

Currently, the pVACtools page lists support for four diff MHC Class II predication algorithms - NetMHCIIpan , SMMalign, NNalign and, MHCnuggets. The first three were added via IEDB, and MHCnuggets was added outside of the IEDB framework.

However, the IEDB webpage lists support for two additional algorithms that I don't see listed on our website- Combinatorial library and Sturniolo. http://tools.iedb.org/mhcii/help/#Method

Are these two not supported by pVACtools? If so, why?

[susannasiebert]

They’re not currently supported but should be trivial to add as long as the IEDB output format is similar to the other algorithms.

[susannasiebert]

Combinatorial library is available standalone using the comblib method. Sturniolo is not listed in this table of available methods on the API: http://tools.iedb.org/main/tools-api/ (MHC-II binding command line parameters: table). Could Sturniolo be the teptiope method? Or is that something else altogether?

[susannasiebert]

According to this: http://tools.iedb.org/auto_bench/mhcii/server/6 teptiope seems to correspond to Sturniolo.

[susannasiebert]

Both Sturniolo and Combinatorial Library output a score instead of a ic50 value. We need to investigate if and how they might correspond to ic50 values. For Sturniolo the scores might even be negative values:

$ curl --data "method=tepitope&sequence_text=SLYNTVATLYCVHQRIDV&allele=HLA-DRB1*01:01" http://tools-cluster-interface.iedb.org/tools_api/mhcii/
allele  seq_num start   end core_peptide    peptide score   rank
HLA-DRB1*01:01  1   1   15  YNTVATLYC   SLYNTVATLYCVHQR -0.1500 4.46
HLA-DRB1*01:01  1   2   16  YNTVATLYC   LYNTVATLYCVHQRI -0.1500 4.46
HLA-DRB1*01:01  1   3   17  YNTVATLYC   YNTVATLYCVHQRID -0.1500 4.46
HLA-DRB1*01:01  1   4   18  VATLYCVHQ   NTVATLYCVHQRIDV -1.0000 8.51

This doesn't fit well into our current paradigm with automatic binding cutoff filtering and the way we aggregate the ic50 scores.

[malachig]

yeah will definitely have to think about this. may have to treat these separately.

[susannasiebert]

I was hoping that with adding the percentile feature we would be able to start supporting these algorithms but they output ranks, not percentiles. I'm removing this issue from the 2.0.0 milestone again.