13-P

Abstract

Aim Here, we describe the web-based tool EpHLA-Converter, a free program designed to reduce the complexity of the problem to predict high resolution HLA typing. Methods Our program allows one to determine the most probable high-resolution HLA class I and II genotypes using as data input just low- or medium-resolution HLA-A, -B, -C, -DRB1, -DRW, -DQA1 and -DQB1 typing and the population it came from. Allelic conversion is based on public databases (NMDP, IMGT) queries and allele frequencies determined from either reference populations or private HLA typing databases. The program was developed to run independently or to be integrated to third part programs including the recently developed EpHLA program to analyze epitopic HLA serum reactivity. We validated the EpHLAConverter to HLA class II inference by using three different approaches: (i) We determined the percentage of matched results between the 100 predicted high-resolution typing with the actual high-resolution typing to HLA-DRB1, (ii) the concordance to predict HLA-DQA1 and HLA–DQB1 assignments by using NMDP published DRB1/DQB1/DQA1 associations and (iii) by computing the time spent in the conversion from medium- to high-resolution HLA- A, -B and -DRB1 typing for database with more than 10,000 typings each one. Results Percentage matched predicted High-Resolution vs. Actual High-resolution typing for HLA-DRB1 and –DQB1 was 94% and 83%, respectively. Besides, we showed that our tool is able to safely convert HLA typing databases with 10,000 typings HLA-A, -B –DRB1 in 191 minutes (more than five inferences per second). The limitation of the EpHLA-Converter program is if the allele DRB1 is uncommon, the association HLA-DRB1/DQA1/DQB1 will not be able to be inferred because the association is not listed on the available frequencies tables. Conclusions Our data demonstrates that EpHLA-Converter has an excelent accuracy rate to predict DRB1 and DRB1/DQA1/DQB1 association at high resolution level.