MARIA (MHC Analyzer with Recurrent Integrated Architecture) is an integrated tool to predict how likely a peptide to be presented by HLA-II complexes on cell surface. MARIA deep neural network learned from over 60,000 naturally presented HLA-II peptide ligands considering their gene expression levels, binding affinities, cleavage signatures, and peptide sequences. Our work showed MARIA better identify presented HLA-II antigens and CD4 T-cell epitopes than conventional tools trained on in vitro binding affinities. MARIA for HLA-I is under active development.
Users should modify the input file template from the Download section. The input file is a plain tap-delimited text file with a header and 5 required columns. Column 1 and 2 are HLA-DR or DQ alleles of the cells (see Supported Alleles for details). Column 3 is the gene symbol (e.g. CTSK) of genes encoding the peptide of interest. Column 4 is peptide sequences in single letter format (all capitalized, no space). Column 5 is optional gene expression values if users want to provide specific gene expression values for this antigen gene (in TPM). Otherwise gene expression values will be estimated from external RNA-Seq references (e.g. TCGA) and genes with unknown gene expression will be assigned with a TPM of 5.
This work was supported by the National Institutes of Health (NIH) U01 CA194389 (M.M.D., R.L., J.E.E., A.A.A.), NIH K08 CA207882 (M.S.K.), NIH GM 102365 (R.B.A.), NIH S10 RR02933801, NIH/Stanford MSTP training grant (B.C.), NSF GSF (E.F.), American Society of Hematology Scholar Award (A.A.A.), V-Foundation (A.A.A.), Damon Runyon-Rachleff Innovation Award (J.E.E.), W.M. Keck Foundation Medical Research Grant (J.E.E.), Conquer Cancer Foundation Young Investigator Award (M.S.K.), Leukemia & Lymphoma Society (M.S.K.), Knut and Alice Wallenberg Foundation Postdoctoral Fellowship (N.O.), Parker Institute for Cancer Immunotherapy Bedside to Benchtop Grant (A.A.A.), PD Soros New American Fellowship (B.C.), Stanford Bio-X Fellowship (B.C.). This work used the XStream computational resource, supported by the National Science Foundation Major Research Instrumentation program (ACI-1429830). This work used the shared FACS facility, supported by NIH S10 Shared Instrument Grant (S10RR027431-01).
Please send questions, issues, and/or licensing requests to: firstname.lastname@example.org