This model based on deep learning method can be used to predict peptide binding MHC class I molecules. The advantage of this model is to ignore the length of peptide. In other words, the same model can be used to predict different peptides with different length. We have trained different models for 66 alleles, including 49 human alleles, three mouse alleles, eight macaque alleles, five chimpanzee alleles and one rat allele.

Step1: Input peptide sequence
Enter peptide sequence(s)

Step2: Select a species and Allele(s)
Allele(s) or
Step3: Set output
Output Value
The “Score” is a self-coined index, taking values between 0 and 1, and it is converted from IC50. 
										The closer the score to 1, the higher likelihood of a binder. 
										The “IC50” measures the binding affinity between a peptide and a MHC protein.

All steps to use the model to predict peptide binding MHC class I are as follow:

  • To paste a peptide sequence into the text area, or to upload a file containing peptide sequences.
  • To select a species.
  • To select allele or to upload a file containing some alleles.
  • To select a output types including ic50 or score.
  • This dataset including 67 allele subdatasets, including 49 hunman alleles and 18 other species alleles. Every dataset contains four files, including training set, test set, validation set and independent test set. We also supply python code of biLSTM models. Please download datasets by clicking the link below:

    Other species
    Independent dataset

    We appreciate your comments to make our website more complete in the future, please contact:

    Yan Guo, Ph.D,
    Limin Jiang, M.S.,