RATE

The RATE (Restrictor Analysis Tool for Epitopes) tool infers HLA restriction for a set of epitopes from large datasets of T cell responses in HLA-typed subjects. It was developed with a focus on class II alleles, but can also be applied to class I alleles.

The tool offers two prediction models. RATE evaluates one allele at a time. COMBO-RATE extends it by evaluating combinations of alleles, because a single epitope is frequently presented by more than one HLA allele.

Methods

Both models take the same two data files: one containing the alleles expressed by the subjects, the other containing the response of each peptide in each subject. They differ in whether they evaluate one allele at a time or combinations of alleles.

RATE

For every peptide-allele combination, RATE builds a matrix tabulating positive and negative responders among the subjects who express that HLA type, against positive and negative responders among the subjects who do not. From that matrix it calculates an odds ratio and a relative frequency, and uses Fisher’s exact test for the significance of each combination. The peptide-allele combinations reported as significant are the inferred HLA restrictions.

References:

  • Paul S, Arlehamn CSL, Schulten V, Westernberg L, Sidney J, Peters B, Sette A. 2017. Experimental validation of the RATE tool for inferring HLA restrictions of T cell epitopes. BMC Immunology 18(1):20. [PMID: 28681704]

  • Paul S, Dillon MBC, Arlehamn CLS, Huang H, McKinney DM, Scriba TJ, Sidney J, Davis MM, Peters B, Sette A. 2015. A Population Response Analysis Approach To Assign Class II HLA-Epitope Restrictions. Journal of Immunology 194(12):6164-6176. [PMID: 25948811]

COMBO-RATE

Single-allele approaches are limited because T cell epitopes are frequently presented by multiple HLA alleles, a phenomenon termed promiscuous restriction. COMBO-RATE extends RATE by systematically evaluating combinations of HLA alleles to identify multi-allelic restriction patterns.

In the study describing it, promiscuous restriction was found to be a near-universal feature of immunodominant epitopes. Applied to the 43 immunodominant CD4 T cell epitopes from a B. pertussis genome-wide screen, COMBO-RATE identified restrictions for 35 of the 43 epitopes compared with 24 by RATE alone, and uncovered 64 additional allele restrictions including 29 unique alleles. The inferred restrictions were confirmed experimentally using single-HLA transfected cell lines and antigen presentation assays.

Reference: Nevarez-Mejia J, Trevizani R, Abawi A, Johansson E, Sutherland A, Grifoni A, da Silva Antunes R, Sette A. 2026. COMBO-RATE: An experimentally validated bioinformatic tool to identify promiscuous HLA restrictions. bioRxiv. [DOI: 10.64898/2026.04.14.718521]

Input Data

RATE Input

RATE requires two tab-delimited plain text files. Both can be generated by copying data out of a spreadsheet.

Allele data

The allele typing data for each subject.

  • First row: donor or subject names, for example Donor-1, Donor-2.

  • Subsequent rows: each row is a different HLA locus, for example DRB1, DRB3, DRB4, DRB5, DQA1, DQB1, DPA1, DPB1.

  • Each cell: the specific allele for that donor at that locus, for example DRB1*03:01.

Both DRB1*03:01 and HLA-DRB1*03:01 are accepted.

Response data

The response of each peptide in each subject.

  • First row: a peptide column header, then the donor names. The header is not case-sensitive, so Peptide is also accepted.

  • Subsequent rows: one row per peptide, with its sequence in the peptide column. Every row needs a sequence, and no sequence can repeat.

  • Response values: either absolute values such as raw assay measurements, or binary 1/0 values indicating a positive or negative response.

Any other column, such as Peptide_ID or Peptide #, is accepted with a warning and dropped before analysis. It does not appear in the results. A file whose sequence column is headed Peptide_Seq is rejected; rename that header to peptide.

The two files must agree

The number and names of the donors must be exactly the same in both files, and donor names are case-sensitive. Both files need the same number of donor columns.

Unknown allele types and unavailable response values are represented as n/a. Missing response values are not included in the statistical calculations.

Parameter Selection

Response Threshold (SFC)

The value at which a response counts as positive.

  • For absolute values, provide the desired cutoff. Values at or above it are converted to 1, values below it to 0.

  • For binary values, provide 1.0, so that all non-zero values count as positive.

The examples work with a cutoff of 20.0, which is the default.

Prediction Model

This control selects which of the two models is run, and it determines what else appears on the form. RATE exposes no further options. COMBO-RATE adds two independent checkboxes.

RATE

Prediction Model set to RATE

Each allele is evaluated against each peptide on its own, and restrictions are reported against single alleles.

COMBO-RATE

Combinations of alleles are evaluated alongside single alleles, and two checkboxes become available.

Pair Chains treats the alpha and beta chains as paired molecules, so restrictions are reported against pairs such as DQA1*05:01/DQB1*05:01.

COMBO-RATE with Pair Chains selected

Predict Binding Affinities runs an MHC binding prediction and uses it to narrow the candidate alleles. The Binding prediction percentile rank cutoff field appears once the box is checked, and only peptide-allele pairs whose eluted ligand percentile rank passes it are carried into the restriction analysis. Loosening the cutoff therefore admits more candidates and lengthens the HLA Restriction table. On the example data, raising it from 2 to 50 takes that table from 59 rows to 187.

Enabling the box also adds a second results tab, MHC elution, reporting the predicted binding of each peptide-allele pair.

COMBO-RATE with binding prediction and a percentile rank cutoff of 2

Selecting both applies paired chains to the restriction analysis and to the binding prediction, so the alleles in the MHC elution tab are reported as pairs as well.

COMBO-RATE with paired chains and a binding prediction cutoff

Results

The results page reports the analysis for every peptide-allele combination, so the table is long: a run of 5 peptides against the 90 alleles expressed across the subjects produces 450 rows.

HLA Restriction

RATE HLA Restriction table

Column Description
peptide The peptide sequence, as given in the response data.
Restricting allele The allele being evaluated against that peptide. A slash joins the two chains of a paired molecule, as in DQA1*05:01/DQB1*05:01. A plus joins the members of an allele combination found by COMBO-RATE, as in DRB1*07:01+DRB1*01:01.
A+R+ Subjects expressing the allele who have a positive response.
A-R+ Subjects who do not express the allele but have a positive response.
A+R- Subjects expressing the allele who do not have a positive response.
A-R- Subjects who neither express the allele nor have a positive response.
Rel. freq. Relative frequency. Used in place of the odds ratio where the odds ratio becomes infinite through division by zero, which happens when no subject lacking the allele responds (A-R+ = 0).
Odds ratio Strength of association between expression of the allele and a positive response. Greater than 1 indicates a positive association.
p-value From Fisher's exact test. Useful as a relative ranking of which restriction has the most statistical support.
Significant Whether the combination is reported as an inferred restriction.

MHC elution

Shown when Predict Binding Affinities is enabled.

RATE MHC elution table

Column Description
peptide The peptide sequence.
allele The allele, or the paired chains when Pair Chains is enabled.
class MHC class.
percentile rank Predicted binding percentile rank. Lower is a better binder.
EL score Eluted ligand score from the same prediction. Higher is a stronger predicted ligand, so it runs in the opposite direction to the percentile rank beside it.
method The prediction method used: netmhcpan_el for class I, netmhciipan_el for class II. Hidden by default; show it with Display Columns.

This tab reports every peptide-allele pair that was scored, so percentile ranks above the Binding prediction percentile rank cutoff appear here alongside those that passed it.

Because method is hidden by default, it is present in an All Rows download whether or not it has been switched on, and absent from a Displayed Rows download unless it has. See Downloads for what each choice carries.

Interpreting Results

The peptide-allele combinations reported as significant are the inferred HLA restrictions. An odds ratio above 1 indicates a positive association between expression of the allele and detection of a positive response, and the p-value ranks which restrictions have the most statistical support.

Where no subject lacking the allele responded, the odds ratio is infinite and the relative frequency should be read in its place.