NetMHCstab - 1.0

Stability of peptide:MHC-I complexes

NetMHCstab-1.0 predicts the stability of peptide binding to a number of different MHC molecules using artificial neural networks (ANNs).

ANNs have been trained for 13 different Human MHC (HLA) alleles representing 11 of the 12 common HLA A and B Supertypes as defined by Lund et al. (2004).

Prediction values are given in hours (half-life).

Predictions of lengths 8-14: Predictions can be made for lengths between 8 and 14 for all alleles using an approximation algorithm using ANNs trained on 9mer peptides. Caution should be taken for 8mer predictions as some alleles might not bind 8mers to any significant extend.

Highly and weakly stable binding peptides are indicated in the output.

The project is a collaboration between CBS and IMMI.


Type of input

Paste a single sequence or several sequences in FASTA format into the field below:

or submit a file in FASTA format directly from your local disk:

Peptide length (several lengths are possible): 

Select loci/species

Select Allele(s) or type allele names (ie HLA-A01:01) separated by commas (and no spaces).

For list of allowed allele names click here List of MHC allele names.

Threshold for highly stable binder (hours):  
Threshold for weakly stable binder (hours):  

Include affinity predictions   Relative weight of affinity prediction: 

Sort by score  

Save prediction to xls file 

At most 5000 sequences per submission; each sequence not more than 20,000 amino acids and not less than 8 amino acids.

The sequences are kept confidential and will be deleted after processing.


For publication of results, please cite:

  • NetMHCstab - predicting stability of peptide:MHC-I complexes; impacts for CTL epitope discovery
    Kasper W. Joergensen, Michael Rasmussen, Soren Buus, and Morten Nielsen
    Immunology. 2013 Aug 8. doi: 10.1111/imm.12160. [Epub ahead of print] 23927693


1. Specify the input sequences

All the input sequences must be in one-letter amino acid code. The allowed alphabet (not case sensitive) is as follows:

A C D E F G H I K L M N P Q R S T V W Y and X (unknown)

All the other symbols will be converted to X before processing.

The server allows for input in either FASTA or PEPTIDE format.

Note that for Peptide input, all peptides MUST of equal length. Note also, that you must click the box Click if input is PEPTIDE format if the input is in peptide format.

The sequences can be input in the following two ways:

  • Paste a single sequence (just the amino acids) or a number of sequences in FASTA format or a list of peptides into the upper window of the main server page.

  • Select a FASTA or PEPTIDE file on your local disk, either by typing the file name into the lower window or by browsing the disk.

Both ways can be employed at the same time: all the specified sequences will be processed. However, there may be not more than 10 sequences in total in one submission. The sequences shorter than 15 or longer than 10000 amino acids will be ignored.

2. Customize your run

Select the allele(s) you want to make predictions for from the scroll-down menu (select multiple alleles using the ctrl key), or type in the allele names separated by commas (with out blank spaces).

Give threshold value for binding values to be displayed.

Click the box Sort by affinity to have the output sorted by descending predicted binding affinity

Click the box save prediction to xls file to save the raw prediction output to an excel file. This file will be available in the bottum of the results output file.

3. Submit the job

Click on the "Submit" button. The status of your job (either 'queued' or 'running') will be displayed and constantly updated until it terminates and the server output appears in the browser window.

At any time during the wait you may enter your e-mail address and simply leave the window. Your job will continue; you will be notified by e-mail when it has terminated. The e-mail message will contain the URL under which the results are stored; they will remain on the server for 24 hours for you to collect them.

Output format


The prediction output consists of 8 columns.

  • Residue number
  • HLA Allele
  • Peptide sequence
  • Protein identifier
  • Prediction score (called 1-log50K(aff))
  • Affinity as IC50 value in nM (only for white-listed alleles)
  • %Random - %Rank of prediction score to a set of 1000.000 random natural 9mer peptides
  • Binding level (SB: strong binder, WB: weak binder)


    # NetMHCpan version 2.8
    # Input is in FSA format
    HLA-A0101 : Estimated prediction accuracy  0.811 (using nearest neighbor HLA-A0101)
    # Threshold for Strong binding peptides  50.000
    # Threshold for Weak binding peptides 500.000
      pos        HLA    peptide        Identity 1-log50k(aff) Affinity(nM)  %Random  BindLevel
        0 HLA-A*0101  ASQKRPSQR seq2_optional_c         0.063     25230.98    32.00
        1 HLA-A*0101  SQKRPSQRH seq2_optional_c         0.023     38824.58    50.00
        2 HLA-A*0101  QKRPSQRHG seq2_optional_c         0.003     48254.07    50.00
        3 HLA-A*0101  KRPSQRHGS seq2_optional_c         0.009     45287.57    50.00

    Article Abstract

    Main reference:

    NetMHCstab - predicting stability of peptide:MHC-I complexes; impacts for CTL epitope discovery
    Kasper W. Joergensen1, Michael Rasmussen3, Buus S3, Nielsen M1,2,
    Immunology. 2013 Aug 8, [Epub ahead of print]

    1Center for Biological Sequence Analysis, Technical University of Denmark, DK-2800 Lyngby, Denmark
    2Instituto de Investigaciones Biotecnologicas, Universidad Nacional de San Martin, San Martin, Buenos Aires, Argentina
    3Division of Experimental Immunology, Institute of Medical Microbiology and Immunology, University of Copenhagen, Denmark

    Major Histocompatibility Complexes (MHC) play an essential role in the cellular immune response, presenting peptides to CTL cell allowing the immune system to scrutinize ongoing intracellular production of proteins. In the early 1990's immunogenicity and stability of the peptide:MHC-I complex were shown to be correlated. Measuring stability was however at that time very cumbersome and time consuming and only small data sets were analyzed. The focus was then turned to binding affinity and stability was put aside. Here, we investigate this fairly unexplored area on a very large scale compared to earlier studies. In a recent publication it was for the HLA-A02:01 system demonstrated that stability was a better predictor than peptide affinity of CTL immunogenicity. Using a similar approach, we here analyzed a total 5,509 distinct peptide stability measurements derived using recently developed high-throughput assay covering 10 different HLA class I molecules. Artificial neural networks were used for constructing novel stability predictors able of predicting the half-life of the peptide:MHC-I complex. These predictors were shown to predict T-cell epitopes and MHC ligands from SYFPEITHI and IEDB to form significantly more stable MHC-I complexes compared to affinity-matched non-epitopes. Further, combining the stability predictions with today's state of the art affinity predictions NetMHCcons (Karosiene et al., 2012) significantly improves the performance for identification of T-cell epitopes and ligands. We identified for most of the HLA alleles included in the study, distinct sub-motifs that differentiates stable from unstable peptide binders, where especially anchor position in the N-terminal of the binding motif (primarily P2 and P3) where found to play a critical role for stable pMHC-I interactions. A webserver implementing the method is available at

    Software Downloads


    Correspondence:        Technical Support: