Closed ndeng1 closed 1 year ago
Hi @ndeng1,
Thank you for the detailed bug report. A PR to fix this issue is now up (#945). I will make a hotfix in the next couple of days.
A hotfix (3.1.2) for this issue is now out. I'm closing this issue but please do reopen it if you continue running into this problem.
Describe the bug When using NNAlign prediction alone in pVACbind, the results are consistent with those obtained from IEDB online prediction. However, incorporating other algorithms such as "all", "class_ii", or "NNAlign MHCNuggetII" leads to significant changes in the NNAlign output.
To Reproduce
Generate a sample file.
Run the NNalign only
Here is the output
Check the NNalign Score of peptide
GGGSEKKKGLMTL
The NNalign Score obtained from predicting with MHC Nuggets II is 10935.8, which matches the result obtained from http://tools.iedb.org/mhci.
Run the NNalign together with MHC Nuggets II.
Log
Check the NNalign Score of peptide
GGGSEKKKGLMTL
The NNalign Score obtained from the prediction with MHC Nuggets II is 13.1. This result remains consistent whether we use the "all" or "class_ii" option.