Outdated HLA Class II models

[riederd]

Thank you for your reply. I’ve been running the pipeline using a sample with both Class I and Class II HLA alleles supplied in the same format as your test dataset (one allele per line, including DR, DQ, and DP). However, I noticed that the canonical Class II neoantigen predictions (from pVACseq) included only DRB1 alleles, whereas other alleles like DPA1/DPB1 were found in the NeoFuse results.

Could you please clarify:

Whether nextNEOpi handles DQ/DP pairing internally when running pVACseq?

If so, is there any filtering or compatibility check that might be excluding certain combinations from showing in(*MHC_Class_II_all_epitopes_ccf_ref_match.tsv)?

What is the recommended format and content for the HLA input file to ensure full Class II coverage across pVACseq and NeoFuse?

Any insight would be greatly appreciated to ensure I’m following best practices.

Originally posted by @nemke333 in #64

@nemke333

Thanks for the catch, we missed updating the class II model file should be fixed now. See assets/hlaii_models.txt for the supported list.

[nemke333]

Thank you again for updating hlaii_models.txt.

I performed a clean reinstallation of the pipeline and ran it using the provided test dataset, supplying the test_data custom_hla.txt file. The Final_HLA calls folder output included multiple Class II alleles, including DRB*, DQA1, DQB1, DPA1, and DPB1.

However in the NeoFuse fusion predictions, only a DP dimer (DPA101:03/DPB102:01) appeared along with several DRB* alleles however — no DQ dimers were reported.

In the canonical pVACseq predictions, only DR alleles (DRB1, DRB3, DRB4, DRB7–9) were included. No DQ or DP predictions appeared, despite the presence of matching alleles in the HLA call output and their inclusion in the updated model file.

Could you kindly confirm whether:

Further configuration is required to enable DQ/DP in canonical pVACseq predictions?

NeoFuse is expected to include both DQ and DP dimers, or if it currently only supports DP?

I’d be happy to provide exact input files or logs if helpful.

[riederd]

Hi, I think we found the root of the issue. We will provide a fix soon, thanks.