New matrix for EpiQuest-T correctly predicts the "strength" of a CTL epitope

EpiQuest has received recently a valuable addition: a new matrix to predict areas containing CTL epitopes. This is our first matrix for mouse CTL epitopes, as all matrixes for EpiQuest-T it is MHC-allele specific, and this one is for the most widely used mouse model, mice with H2kB haplotype.

Obviously, everyone is interested not in any CTL epitope, but in a "strong" one, as this property of the peptide sequence defines how effective will be CTLs against their target.

To investigate whether our algorithm (and matrix) for H2kB recognises the quality of CTL epitope correctly, we performed a retrospective study, using H2kB epitopes from the IEDB ( database that were described as "strong", "ïntermediate" or "weak/negative" in functional tests. We have analysed all epitopes from the first two categories, and the random selection of the negative ones. A random selection of the strong peptide epitopes for another mouse allele, H2kD were included in a study to investigate the allele restriction of the epitope strength recognition.

All epitopes were analysed using EpiQuest-T, and the AGI score for the best 8-mer from a given epitope was recorded for all epitopes. The resulting scores are presented in an image below:

The main conclusion from the observed range of AGI values for each group of epitopes was that the algorithm correctly discriminates between functionally active and non-active epitopes, and that the recognition is allele-specific.

All 4 groups are significantly different, which confirmed statistically (medians were compared):

Comparison of the groups in pairs revealed that strong (highly positive) epitopes differ in their AGI highly significant from both intermediate and negative ones, as well as from strong (high) epitopes of another allele.

On the overall, considering that definitions of strong and weak/negative epitopes, as used in IEDB, are not very strict, and within the limitations of the study, the new algorithm and matrix for H2kB epitopes is highly accurate.

Featured Posts
Recent Posts
Search By Tags