Predivac is a method to predict HLA class II peptide binding, based on the specificity-determining residue (SDR) concept, which covers 95% of MHC class II allelic variants (DR locus). It is implemented as a computational framework that integrates prediction of CD4+ T-cell epitopes in proteins (antigens) with population coverage and epitope selection algorithms to optimise the selection of multiple putative epitopes targeting well defined ethnic populations. The population coverage assessment makes use of "the Allele Frequency Net Database" (AFND), which is the most comprehensive repository of immune gene frequencies of worldwide population. The comprehensiveness of the tool in terms of HLA class II allele coverage and allele frequencies makes it an optimal tool to aid epitope-based vaccine design in the context of a genetically heterogeneous human population.
Peptide binding prediction: The program evaluates a single peptide or a peptide list for binding prediction on a given HLA class II protein (DRB locus) set by the user. For every peptide sequence, the program assesses all possible nonameric sequences and assigns a score (0-100) to each. The peptide score is assigned to the nonamer having the highest score (corresponding to the binding core). The peptides must be listed on a flat (text) file, either in fasta format or as a simple list.
Protein binding prediction: The program predicts putative CD4+ T-cell epitopes in a given protein sequence (submitted in fasta format), for one specific MHC class II allele. Epitope mapping is carried out by parsing query protein sequences into overlapping nonameric segments (peptides), each of which is assigned a binding score (using the sliding windows technique). The higher the score, the higher the probability that a peptide is an HLA class II high affinity binder and, potentially, a CD4+ T-cell epitope. The default threshold value is 3%, as most immunodiominant epitopes occur within this range, though Predivac allows selecting peptides with thresholds ranging between 1% and 5% of the top scores.
Binding prediction is carried out with the aim of identifying putative CD4+ T-cell epitopes with a potential application in rational vaccine design.
Population coverage: CD4+ T-cell epitopes are predicted in proteins (antigens), not for a specific MHC class II alleliec variant (which is performed using the binding prediction option), but for a largeset of HLA class II proteins occuring in a given human population (target population), associated to a given geographyc region, country or single ethnicity, according to the Allele Frequency Net DB. The population coverage is calculated thrugh a process that combines the predicted epitopes using either a simplified method (simple search) or a more sophisitcated optimization algorithm (optimised search):
Simple search: Population coverage is obtained by summing up the coverage associated with each epitope listed in a sorted list (maximal to minimal coverage), descending through it until the cumulative coverage no longer increases. It provides a quick/on-line response. This is the default mode.
Optimised search: The program implements a genetic algorithm that enables to explore in detail the epitope domain, finding numerous peptide combinations (solutions) that maximise coverage in the target population. This search is slower and the outcome is sent back to the user by e-mail.
Validated epitopes: As a last step of the pipeline, this option allows to calculate the population coverage based on experimentally characterised epitopes plus their allele restriction (HLA class II proteins). The information must be submitted in a flat (text) file, containing the epitope sequence followed by the allele restriction (csv format).