A Hidden Markov Model Combining Linkage and Linkage Disequilibrium Information for Haplotype Reconstruction and Quantitative Trait Locus Fine Mapping

Abstract

Faithful reconstruction of haplotypes from diploid marker data (phasing) is important for many kinds of genetic analyses, including mapping of trait loci, prediction of genomic breeding values, and identification of signatures of selection. In human genetics, phasing most often exploits population information (linkage disequilibrium), while in animal genetics the primary source of information is familial (Mendelian segregation and linkage). We herein develop and evaluate a method that simultaneously exploits both sources of information. It builds on hidden Markov models that were initially developed to exploit population information only. We demonstrate that the approach improves the accuracy of allele phasing as well as imputation of missing genotypes. Reconstructed haplotypes are assigned to hidden states that are shown to correspond to clusters of genealogically related chromosomes. We show that these cluster states can directly be used to fine map QTL. The method is computationally effective at handling large data sets based on high-density SNP panels. Present-day genotyping platforms do not directly provide information about linkage phase; i.e., co-inherited alleles at adjacent heterozygous markers (haplotypes) are not identified as such. As haplotype information may considerably empower genetic analyses, indirect phasing strategies have been devised: haplotypes can be reconstructed from unphased genotypes using either familial information (Mendelian segregation and linkage) and/ or population information (linkage disequilibrium, LD, and surrogate parents) (e.g., Windig and Meuwissen 2004;Scheet and Stephens 2006;Kong et al. 2008).Haplotype-based approaches are routinely applied in animal genetics for combined linkage and LD mapping of QTL (e.g., Meuwissen and Goddard 2000;Blott et al. 2003). In these studies, phasing has so far relied on familial information provided by the extended pedigrees typical of livestock (e.g., Windig and Meuwissen 2004). This approach, however, leaves a nonnegligible proportion of genotypes unphased, especially for the less connected individuals. After phasing, identity-bydescent (IBD) probabilities conditional on haplotype data-needed for QTL mapping-are computed for all chromosome pairs, using familial as well as population information (hence combined linkage and LD mapping -L 1 LD) (e.g., . However, the use of high-density SNP chips and the analysis of ever larger cohorts render the computation of pairwise IBD probabilities a bottleneck.We herein propose a more efficient, heuristic approach based on hidden Markov models (HMM). It simultaneously phases and sorts haplotypes in clusters that can be used directly for mapping or other purposes. The proposed method exploits familial as well as population information, and imputes missing genotypes. We herein describe the accuracy of the proposed method and its use for L 1 LD mapping of QTL. MATERIALS AND METHODSHaplotype reconstruction and clustering: Due to systematic recording of familial relationship in domestic animals, i...

Identification of loci affecting teat number by genome-wide association studies on three pig populations

“…The 60K SNP data of the 928 F 2 pigs and their F 0 and F 1 ancestors (reference animals) were further employed to impute the 60K SNP genotypes of 815 F 2 pigs (target animals) that were not genotyped by 60K SNP chip by using the following approach [8]. Firstly, haplotypes of the reference and target animals were partially reconstructed based on linkage and Mendelian segregation rules with the 183 microsatellite data using DualPHASE [9]. Secondly, haplotypes of the reference individuals were fully reconstructed with microsatellite data and 60K SNPs data using DAGPHASE, which was iteratively called by Beagle [9].…”

“…Firstly, haplotypes of the reference and target animals were partially reconstructed based on linkage and Mendelian segregation rules with the 183 microsatellite data using DualPHASE [9]. Secondly, haplotypes of the reference individuals were fully reconstructed with microsatellite data and 60K SNPs data using DAGPHASE, which was iteratively called by Beagle [9]. Lastly, missing 60K SNP genotypes of the target pigs were filled via CHROMIBD, in which the linkage and a Markov model were utilized to estimate identity-by-descent (IBD) probabilities between target and parent chromosomes from the genotyped ancestors [10].…”

Toward genomic prediction from whole-genome sequence data: impact of sequencing design on genotype imputation and accuracy of predictions

“…To define haplotypes, we use the 1000 SNPs from the panel within each genome (of five chromosomes 10 MB in size) to assign genome segments to ancestral haplotypes at every marker position, using DualPHASE from the PHASEBOOK package (Druet and Georges, 2010). In DualPHASE, ancestral haplotypes are based on a hidden Markov model, which assigns at each marker position a chromosome segment to an ancestral haplotype.…”

Use of partial least squares regression to predict single nucleotide polymorphism marker genotypes when some animals are genotyped with a low-density panel

“…In a recent study carried out with the bovine 54 K SNP, Weigel et al (2010), using the algorithm implemented in fastPHASE 1.2 software (University of Washington TechTransfer Digital Ventures Program, Seattle, WA, USA), reported a proportion of correctly reconstructed missing SNP of about 0.88% when 90% SNP were predicted. Druet and Georges (2010) combined fastPHASE and Beagle (Browning and Browning, 2007) algorithms to take into account both population (linkage disequilibrium) and familial (Mendelian segregation and linkage) information to predict missing genotypes. They found, with 50% missing genotypes, an imputation error of 3% and 1%, respectively, for sparse and dense marker maps.…”