Implementation of Bayesian methods to identify SNP and haplotype regions with transmission ratio distortion across the whole genome: TRDscan v.1.0

摘要

Realized deviations from the expected Mendelianinheritance of alleles from heterozygous parents havebeen previously reported in a broad range of organisms(i.e., transmission ratio distortion; TRD). Variousbiological mechanisms affecting gametes, embryos,fetuses, or even postnatal offspring can produce patternsof TRD. However, knowledge about its prevalenceand potential causes in livestock species is still scarce.Specific Bayesian models have been recently developedfor the analyses of TRD for biallelic loci, which accommodateda wide range of population structures,enabling TRD investigation in livestock populations.The parameterization of these models is flexible andallows the study of overall (parent-unspecific) TRD andsire- and dam-specific TRD. This research aimed atderiving Bayesian models for fitting TRD on the basisof haplotypes, testing the models for both haplotypeandSNP-based methods in simulated data and actualHolstein genotypes, and developing a specific softwarefor TRD analyses. Results obtained on simulated datasets showed that the statistical power of the analysisincreased with sample size of trios (n), proportion ofheterozygous parents, and the magnitude of the TRD.On the other hand, the statistical power to detectTRD decreased with the number of alleles at each loci.Bayesian analyses showed a strong Pearson correlationcoefficient (≥0.97) between simulated and estimatedTRD that reached the significance level of Bayes factor≥10 for both single-marker and haplotype analyseswhen n ≥ 25. Moreover, the accuracy in terms of themean absolute error decreased with the increase of thesample size and increased with the number of allelesat each loci. Using real data (55,732 genotypes of Holsteintrios), SNP- and haplotype-based distortions weredetected with overall TRD, sire-TRD, or dam-TRD,showing different magnitudes of TRD and statisticalrelevance. Additionally, the haplotype-based methodshowed more ability to capture TRD compared withindividual SNP. To discard possible random TRD inreal data, an approximate empirical null distribution ofTRD was developed. The program TRDscan v.1.0 waswritten in Fortran 2008 language and provides a powerfulstatistical tool to scan for TRD regions across thewhole genome. This developed program is freely availableat http: / / www .casellas .info/ files/ TRDscan .zip.