About

This site gives information on MHC Class I alleles as predicted by MHCflurry.

The results are predictions, not experimentally-determined motifs. In some cases these will deviate from actual motifs. If you find a case of egregious mismatch, please let us know by filing an issue. This will help us improve MHCflurry in the future.

Result generation

We are using the MHCflurry binding affinity (BA) prediction to generate these, not the presentation scores. This may change in the future.

For each allele, we took the top 1% predicted binders across a large set of random peptides. The peptides were sampled using the amino acid frequencies of the peptides in the MHCflurry training data.

For the sequence logos, we plotted the frequency of each amino acid after normalizing by the overall rate of that amino acid in the training data.

The length distributions represent the fraction of the top 1% peptides that are each length.

Code

The predictions over the big pool of peptides are generated by the MHCflurry calibrate-percentile-ranks command.

The motif images and related analyses are part of the analysis_predictor_info MHCflurry download.

This site is generated in the mhcflurry-motifs GitHub project.

Citation

If you find these results useful in your research, please cite MHCflurry:

T. O'Donnell, A. Rubinsteyn, U. Laserson. "MHCflurry 2.0: Improved pan-allele prediction of MHC I-presented peptides by incorporating antigen processing," Cell Systems, 2020. https://doi.org/10.1016/j.cels.2020.06.010

T. O’Donnell, A. Rubinsteyn, M. Bonsack, A. B. Riemer, U. Laserson, and J. Hammerbacher, "MHCflurry: Open-Source Class I MHC Binding Affinity Prediction," Cell Systems, 2018. https://doi.org/10.1016/j.cels.2018.05.014