How to interpret

How to understand the output

After converting the .csv-file into a spreadsheet (see How to use) it should look like this:


We recommend transforming the data field into a table to easily sort for best predictions.

The columns contain the following information:

Position Start: the position of the first amino acid of the predicted peptide within the input sequence

Region: the start and end position of the predicted peptide

Length: the length of the predicted peptide

Allele: if multiple alleles were selected for prediction this column contains the respective HLA-allele

Sequence: the amino acid sequence of the predicted peptide

Algorithm Name: the output values of the respective method

"N/A" means that a peptide is not predicted by the server
e.g. if the server does not perform prediction of the chosen peptide length.


The output values of the prediction methods are:

IC50 in nM (NetMHC methods, IEDB smm and IEDB smmpmbec):
The lower the value, the stronger the binding affinity.
Usual thresholds are IC50 ≤ 50nM (strong affinity) and IC50 ≤ 500nM (low affinity).

Percentile rank (NetMHC methods, IEDB recommended and IEDB consensus):
The lower the value, the higher the chances of binding.
Usual thresholds are %-rank ≤ 0.5 (strong affinity) and %-rank ≤ 2 (low affinity).
NetMHCpan 4.1 differentiates %-rank prediction based on either LC-MS eluted ligands (EL) or binding affinity (BA).

Score (SYFPEITHI):
The higher the binding score, the higher the chances of binding.
SYFPEITHI does not recommend any threshold.

Please see Recommendations for information about increasing prediction sensitivity by using more tolerant thresholds.