Predicting HLA genotypes using unphased and flanking single-nucleotide polymorphisms in Han Chinese population
- Equal contributors
1 Institute of Biomedical Sciences, Academia Sinica, Nankang, Taipei, Taiwan
2 Clinical Informatics and Medical Statistics Research Center, Chang Gung University College of Medicine, Taoyuan, Taiwan
3 Department of Medical Genetics, National Taiwan University Hospital, Taipei, Taiwan
4 Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan
5 Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan
6 Department of Internal Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan
7 National Genotyping Center, Academia Sinica, Taipei, Taiwan
8 Graduate Institute of Chinese Medical Science, China Medical University, Taichung, Taiwan
9 Department of Pediatrics, Duke University Medical Center, Durham, North Carolina, USA
10 Immunogenetics laboratory, Medical Research Department, Mackay Memorial Hospital, Taipei, Taiwan
11 Research Center for Developmental Biology and Regenerative Medicine, National Taiwan University, Taipei, Taiwan
12 Graduate Institute of Medical Genomics and Proteomics, National Taiwan University College of Medicine, Taipei, Taiwan
13 Graduate Institute of Biostatistics, China Medical University, Taichung, Taiwan
BMC Genomics 2014, 15:81 doi:10.1186/1471-2164-15-81Published: 29 January 2014
Genetic variation associated with human leukocyte antigen (HLA) genes has immunological functions and is associated with autoimmune diseases. To date, large-scale studies involving classical HLA genes have been limited by time-consuming and expensive HLA-typing technologies. To reduce these costs, single-nucleotide polymorphisms (SNPs) have been used to predict HLA-allele types. Although HLA allelic distributions differ among populations, most prediction model of HLA genes are based on Caucasian samples, with few reported studies involving non-Caucasians.
Our sample consisted of 437 Han Chinese with Affymetrix 5.0 and Illumina 550 K SNPs, of whom 214 also had data on Affymetrix 6.0 SNPs. All individuals had HLA typings at a 4-digit resolution. Using these data, we have built prediction model of HLA genes that are specific for a Han Chinese population. To optimize our prediction model of HLA genes, we analyzed a number of critical parameters, including flanking-region size, genotyping platform, and imputation. Predictive accuracies generally increased both with sample size and SNP density.
SNP data from the HapMap Project are about five times more dense than commercially available genotype chip data. Using chips to genotype our samples, however, only reduced the accuracy of our HLA predictions by only ~3%, while saving a great deal of time and expense. We demonstrated that classical HLA alleles can be predicted from SNP genotype data with a high level of accuracy (80.37% (HLA-B) ~95.79% (HLA-DQB1)) in a Han Chinese population. This finding offers new opportunities for researchers in obtaining HLA genotypes via prediction using their already existing chip datasets. Since the genetic variation structure (e.g. SNP, HLA, Linkage disequilibrium) is different between Han Chinese and Caucasians, and has strong impact in building prediction models for HLA genes, our findings emphasize the importance of building ethnic-specific models when analyzing human populations.