HIBAG--HLA genotype imputation with attribute bagging.

Zheng X, Shen J, Cox C, Wakefield JC, Ehm MG, Nelson MR, Weir BS. · 2014

Read HIBAG--HLA genotype imputation with attribute bagging. by Zheng X, Shen J, Cox C, Wakefield JC, Ehm MG, Nelson MR, Weir BS. (2014) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3772955.txt.

Introduction

X. Zheng and six co-authors presented a method called HIBAG in The Pharmacogenomics Journal in 2014 for predicting HLA types from SNP data. HLA genes guide the immune system's presentation of antigens and matter for disease and drug reaction studies, but direct typing is too costly for large samples. HIBAG predicts HLA alleles from the dense SNP genotypes many studies already hold.

The method averages predictions over many classifiers, each built on a resampled version of the training data. Trained on 2,668 people of European ancestry and tested on about 1,000 people from a British birth cohort, it predicted the five main HLA genes with about 92 to 98 percent accuracy. It matched or beat the two leading rival methods. The free R package ships with ready-made classifiers for European, Asian, Hispanic, and African ancestries.

Catalog record

CREATORZheng X, Shen J, Cox C, Wakefield JC, Ehm MG, Nelson MR, Weir BS.
YEAR2014
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3772955

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

HIBAG—HLA genotype imputation with attribute bagging By X Zheng, J Shen, C Cox, J C Wakefield, M G Ehm, M R Nelson, B S Weir Published in: The Pharmacogenomics Journal DOI: 10.1038/tpj.2013.18 ABSTRACT Genotyping of classical human leukocyte antigen (HLA) alleles is an essential tool in the analysis of diseases and adverse drug reactions with associations mapping to the major histocompatibility complex (MHC). However, deriving high-resolution HLA types subsequent to whole-genome single-nucleotide polymorphism (SNP) typing or sequencing is often cost prohibitive for large samples. An alternative approach takes advantage of the extended haplotype structure within the MHC to predict HLA alleles using dense SNP genotypes, such as those available from genome-wide SNP panels. Current methods for HLA imputation are difficult to apply or may require the user to have access to large training data sets with SNP and HLA types. We propose HIBAG, HLA Imputation using attribute BAGging, that makes predictions by averaging HLA-type posterior probabilities over an ensemble of classifiers built on bootstrap samples. We assess the performance of HIBAG using our study data (n=2668 subjects of European ancestry) as a training set and HLA data from the British 1958 birth cohort study (n≈1000 subjects) as independent validation samples. Prediction accuracies for HLA-A, B, C, DRB1 and DQB1 range from 92.2% to 98.1% using a set of SNP markers common to the Illumina 1M Duo, OmniQuad, OmniExpress, 660K and 550K platforms. HIBAG performed well compared with the other two leading methods, HLA*IMP and BEAGLE. This method is implemented in a freely available HIBAG R package that includes pre-fit classifiers for European, Asian, Hispanic and African ancestries, providing a readily available imputation approach without the need to have access to large training data sets. INTRODUCTION The human leukocyte antigen (HLA) system, located in the major histocompatibility complex (MHC) on chromosome 6p21.3, is highly polymorphic. This region has been shown to be important in human disease, adverse drug reactions and organ transplantation.1 HLA genes have a role in the immune system and autoimmunity as they are central to the presentation of antigens for recognition by T cells. As they have to provide defense against a great diversity of environmental microbes, HLA genes must be able to present a wide range of peptides. Evolutionary pressure at these loci has given rise to a great deal of functional diversity. For example, the HLA-B locus has 1898 four-digit alleles listed in the April 2012 release of the IMGT-HLA Database2 (http://www.ebi.ac.uk/imgt/hla/). Classical HLA genotyping methodologies have been predominantly developed for tissue typing purposes, with sequence-based typing (SBT) approaches currently considered the gold standard. Although there is widespread availability of vendors offering HLA genotyping services, the complexities involved in performing this to the standard required for diagnostic purposes make using a SBT approach time-consuming and cost-prohibitive for most research studies wishing to look in detail at the involvement of classical HLA genes in disease. Previous studies have suggested that the existence of some HLA alleles can be predicted by a single-nucleotide polymorphism (SNP)-based tagging approach.3, 4 However, SNP-based tagging does not offer a definitive solution to HLA genotyping by prediction as many HLA alleles are found on multiple haplotype backgrounds5 that differ among populations. An alternative to tagging is to use more SNP information to impute HLA types. Multiple methods have been developed for this problem, including LDMhc,5, 6 as well as applying general genotype imputation methods such as BEAGLE.7 To be effective, these methods require access to a large and ethnically diverse training data set with both SNP and HLA alleles genotyped. To impute HLA types from multiple SNP markers, Leslie et al.5 used an identity-by-descent model based on approximate coalescent models8 to develop their LDMhc algorithm, and used a leave-one-out cross-validation scheme for SNP selection. Dilthey et al.6 subsequently developed integrated software HLA*IMP for imputing classical HLA alleles from SNP genotypes based on LDMhc, with a modified SNP selection function that leads to pronounced increases in call rate. A training set of SNP haplotypes with known HLA alleles are required by LDMhc, as well as a fine genetic map of the region,6 whereas most experimental techniques for determining SNPs provide genotypes rather than haplotypes. Inferring haplotypes from genotypes can be done with the statistical method of approximating coalescent models, PHASE,9 or newer algorithms like fastPHASE,10 MACH11 and IMPUTE2.12 BEAGLE, an alternative imputation method to the approximate coalescent approach, allows for the prediction of multiallelic loci.7 It locally clusters the observed haplotypes at each position, based on similarity of the haplotypes at markers in the local vicinity.13 It is a computationally efficient approach with high accuracy for thousands of samples and markers. Recently, it was used for HLA imputation in a genetic association analysis.14 That study illustrates how imputation of functional variation can help fine-map association signals in the MHC. Here we propose a new method for HLA Imputation using attribute BAGging, HIBAG, that is highly accurate, computationally tractable and can be used with published parameter estimates, eliminating the need to access large training samples. It combines the concepts of attribute bagging with haplotype inference from unphased SNPs and HLA types. Attribute bagging is a technique for improving the accuracy and stability of classifier ensembles deduced using bootstrap aggregating and random subsets of variables,15, 16, 17 as shown in Figure 1. In this case, individual classifiers are created that utilize a subset of SNPs to predict HLA types and haplotype frequencies estimated from a training data set of SNPs and HLA types. Each of the classifiers employs a variable selection algorithm with a random component to select a subset of the SNPs. HLA-type predictions are determined by maximizing the average posterior probabilities from all classifiers. Compared with LDMhc and BEAGLE, HIBAG has only the minimal assumption of Hardy–Weinberg equilibrium (HWE). We investigate the overall performance of HIBAG using HLA types and SNP genotypes from HapMap, the British 1958 birth cohort data of the Wellcome Trust Case Control Consortium (WTCCC) and HLARES data from GlaxoSmithKline (GSK) clinical trials. We compare HIBAG with two leading methods, HAP*IMP and BEAGLE v3.3. We provide parameter estimates based on our HLA data and software, implementing our method in the freely available HIBAG R package. MATERIALS AND METHODS The numbers of individuals with available four-digit HLA types and the numbers of observed HLA alleles are summarized in Table 1 for the HapMap, WTCCC and HLARES, respectively. Note that sample sizes vary among HLA loci due to missing data. Descriptions of these data follow. HapMap data The HapMap Phase 2 SNP data set consists of (1) 30 parent–offspring trios of Yoruban ancestry from Ibadan in Nigeria, YRI; (2) 30 CEPH trios of European ancestry from Utah, CEU; (3) 45 unrelated Han Chinese from Beijing, CHB; and (4) 45 unrelated individuals from Tokyo in Japan, JPT. The HapMap SNP genotypes (release #28) were downloaded from http://hapmap.ncbi.nlm.nih.gov/downloads/genotypes/2010-08_phaseII+III/forward/. The data set was created by combining genotyping data from several platforms: Affymetrix, Illumina, Perlegen, and so on. When Mendelian errors were detected in a trio, all genotypes for that SNP in that trio were set to missing. SNP markers were selected within the extended MHC (xMHC)18 on chromosome 6 ranging from 025759242 to 033534827 bp. With a missing call rate threshold of 10%, there were 16

Read the complete text