Releases · GfellerLab/MixMHC2pred

15 Mar 12:56

Official release of MixMHC2pred-2.0, a pan-allele predictor of MHC-II ligands and CD4+T-cell epitopes.

Release notes are available here.

Current release corresponds to version 2.0.2 (re-tagged here in order to obtain a permanent Zenodo doi from this release).

Assets 3

03 Oct 10:08

Release of MixMHC2pred-2.0, a major update of our predictor:

This new version allows for pan-allele predictions for human and non-human MHC-II alleles. I.e. it now manages to predict the peptide presentation for alleles that are absent from the training data.
The implementation vastly changed; it is now based on multiple neural networks and it was trained on a much larger MHC-II peptidomics dataset than v1.xx.
Please note that the format of the input and output files have changed, as well as the available input arguments (please check the README for details of the input/output files and usage).

Assets 3

11 Jul 09:56

Pre-release

Pre-release version of the major update of MixMHC2pred-2.0.

This new version allows for pan-allele predictions for human and non-human MHC-II alleles.
The implementation vastly changed; it is now based on multiple neural networks and it was trained on a much larger MHC-II peptidomics dataset.
Please note that the format of the input and output files have changed, as well as the available input arguments (see README for details of the input/output files and usage).

Assets 2

03 Feb 10:21

New version outputting predictions from each allele and various other changes:

Output contains now also the individual predictions from each allele in
addition to the score of the best allele per peptide.
Outputted core positions correspond now to the P1 of the binding core, counted
from the N-terminus of the peptide (first amino acid has a value of 1).
Added a "FAQ" document.
There is an option to allow having unspecified amino acids in the peptide
sequence (these should be written as "X" or "-").
Added the option "--best_s" (see README).
Slightly updated the %Rank computations, avoiding in general having scores of
exactly 0, and allowing the predictions for peptides up to a size of 25 amino
acids.
Corrected a bug when input file was given in fasta format with multiple lines
per peptide sequence.