Mehar S. Khatkar

A comparison of five methods to predict genomic breeding values of dairy bulls from genome-wide SNP markers

BackgroundGenomic selection (GS) uses molecular breeding values (MBV) derived from dense markers across the entire genome for selection of young animals. The accuracy of MBV prediction is important for a successful application of GS. Recently, several methods have been proposed to estimate MBV. Initial simulation studies have shown that these methods can accurately predict MBV. In this study we compared the accuracies and possible bias of five different regression methods in an empirical application in dairy cattle.MethodsGenotypes of 7,372 SNP and highly accurate EBV of 1,945 dairy bulls were used to predict MBV for protein percentage (PPT) and a profit index (Australian Selection Index, ASI). Marker effects were estimated by least squares regression (FR-LS), Bayesian regression (Bayes-R), random regression best linear unbiased prediction (RR-BLUP), partial least squares regression (PLSR) and nonparametric support vector regression (SVR) in a training set of 1,239 bulls. Accuracy and bias of MBV prediction were calculated from cross-validation of the training set and tested against a test team of 706 young bulls.ResultsFor both traits, FR-LS using a subset of SNP was significantly less accurate than all other methods which used all SNP. Accuracies obtained by Bayes-R, RR-BLUP, PLSR and SVR were very similar for ASI (0.39-0.45) and for PPT (0.55-0.61). Overall, SVR gave the highest accuracy.All methods resulted in biased MBV predictions for ASI, for PPT only RR-BLUP and SVR predictions were unbiased. A significant decrease in accuracy of prediction of ASI was seen in young test cohorts of bulls compared to the accuracy derived from cross-validation of the training set. This reduction was not apparent for PPT. Combining MBV predictions with pedigree based predictions gave 1.05 - 1.34 times higher accuracies compared to predictions based on pedigree alone. Some methods have largely different computational requirements, with PLSR and RR-BLUP requiring the least computing time.ConclusionsThe four methods which use information from all SNP namely RR-BLUP, Bayes-R, PLSR and SVR generate similar accuracies of MBV prediction for genomic selection, and their use in the selection of immediate future generations in dairy cattle will be comparable. The use of FR-LS in genomic selection is not recommended.

A primary assembly of a bovine haplotype block map based on a 15,036-single-nucleotide polymorphism panel genotyped in holstein-friesian cattle

Analysis of data on 1000 Holstein–Friesian bulls genotyped for 15,036 single-nucleotide polymorphisms (SNPs) has enabled genomewide identification of haplotype blocks and tag SNPs. A final subset of 9195 SNPs in Hardy–Weinberg equilibrium and mapped on autosomes on the bovine sequence assembly (release Btau 3.1) was used in this study. The average intermarker spacing was 251.8 kb. The average minor allele frequency (MAF) was 0.29 (0.05–0.5). Following recent precedents in human HapMap studies, a haplotype block was defined where 95% of combinations of SNPs within a region are in very high linkage disequilibrium. A total of 727 haplotype blocks consisting of ≥3 SNPs were identified. The average block length was 69.7 ± 7.7 kb, which is ∼5–10 times larger than in humans. These blocks comprised a total of 2964 SNPs and covered 50,638 kb of the sequence map, which constitutes 2.18% of the length of all autosomes. A set of tag SNPs, which will be useful for further fine-mapping studies, has been identified. Overall, the results suggest that as many as 75,000–100,000 tag SNPs would be needed to track all important haplotype blocks in the bovine genome. This would require ∼250,000 SNPs in the discovery phase.

Accuracy of direct genomic values in Holstein bulls and cows using subsets of SNP markers

BackgroundAt the current price, the use of high-density single nucleotide polymorphisms (SNP) genotyping assays in genomic selection of dairy cattle is limited to applications involving elite sires and dams. The objective of this study was to evaluate the use of low-density assays to predict direct genomic value (DGV) on five milk production traits, an overall conformation trait, a survival index, and two profit index traits (APR, ASI).MethodsDense SNP genotypes were available for 42,576 SNP for 2,114 Holstein bulls and 510 cows. A subset of 1,847 bulls born between 1955 and 2004 was used as a training set to fit models with various sets of pre-selected SNP. A group of 297 bulls born between 2001 and 2004 and all cows born between 1992 and 2004 were used to evaluate the accuracy of DGV prediction. Ridge regression (RR) and partial least squares regression (PLSR) were used to derive prediction equations and to rank SNP based on the absolute value of the regression coefficients. Four alternative strategies were applied to select subset of SNP, namely: subsets of the highest ranked SNP for each individual trait, or a single subset of evenly spaced SNP, where SNP were selected based on their rank for ASI, APR or minor allele frequency within intervals of approximately equal length.ResultsRR and PLSR performed very similarly to predict DGV, with PLSR performing better for low-density assays and RR for higher-density SNP sets. When using all SNP, DGV predictions for production traits, which have a higher heritability, were more accurate (0.52-0.64) than for survival (0.19-0.20), which has a low heritability. The gain in accuracy using subsets that included the highest ranked SNP for each trait was marginal (5-6%) over a common set of evenly spaced SNP when at least 3,000 SNP were used. Subsets containing 3,000 SNP provided more than 90% of the accuracy that could be achieved with a high-density assay for cows, and 80% of the high-density assay for young bulls.ConclusionsAccurate genomic evaluation of the broader bull and cow population can be achieved with a single genotyping assays containing ~ 3,000 to 5,000 evenly spaced SNP.

Strategies and utility of imputed SNP genotypes for genomic analysis in dairy cattle

BackgroundWe investigated strategies and factors affecting accuracy of imputing genotypes from lower-density SNP panels (Illumina 3K, 7K, Affymetrix 15K and 25K, and evenly spaced subsets) up to one medium (Illumina 50K) and one high-density (Illumina 800K) SNP panel. We also evaluated the utility of imputed genotypes on the accuracy of genomic selection using Australian Holstein-Friesian cattle data from 2727 and 845 animals genotyped with 50K and 800K SNP chip, respectively. Animals were divided into reference and test sets (genotyped with higher and lower density SNP panels, respectively) for evaluating the accuracies of imputation. For the accuracy of genomic selection, a comparison of direct genetic values (DGV) was made by dividing the data into training and validation sets under a range of imputation scenarios.ResultsOf the three methods compared for imputation, IMPUTE2 outperformed Beagle and fastPhase for almost all scenarios. Higher SNP densities in the test animals, larger reference sets and higher relatedness between test and reference animals increased the accuracy of imputation. 50K specific genotypes were imputed with moderate allelic error rates from 15K (2.85%) and 25K (2.75%) genotypes. Using IMPUTE2, SNP genotypes up to 800K were imputed with low allelic error rate (0.79% genome-wide) from 50K genotypes, and with moderate error rate from 3K (4.78%) and 7K (2.00%) genotypes. The error rate of imputing up to 800K from 3K or 7K was further reduced when an additional middle tier of 50K genotypes was incorporated in a 3-tiered framework. Accuracies of DGV for five production traits using imputed 50K genotypes were close to those obtained with the actual 50K genotypes and higher compared to using 3K or 7K genotypes. The loss in accuracy of DGV was small when most of the training animals also had imputed (50K) genotypes. Additional gains in DGV accuracies were small when SNP densities increased from 50K to imputed 800K.ConclusionPopulation-based genotype imputation can be used to predict and combine genotypes from different low, medium and high-density SNP chips with a high level of accuracy. Imputing genotypes from low-density SNP panels to at least 50K SNP density increases the accuracy of genomic selection.

A First-Generation Metric Linkage Disequilibrium Map of Bovine Chromosome 6

We constructed a metric linkage disequilibrium (LD) map of bovine chromosome 6 (BTA6) on the basis of data from 220 SNPs genotyped on 433 Australian dairy bulls. This metric LD map has distances in LD units (LDUs) that are analogous to centimorgans in linkage maps. The LD map of BTA6 has a total length of 8.9 LDUs. Within the LD map, regions of high LD (represented as blocks) and regions of low LD (steps) are observed, when plotted against the integrated map in kilobases. At the most stringent block definition, namely a set of loci with zero LDU increase over the span of these markers, BTA6 comprises 40 blocks, accounting for 41% of the chromosome. At a slightly lower stringency of block definition (a set of loci covering a maximum of 0.2 LDUs on the LD map), up to 81% of BTA6 is spanned by 46 blocks and with 13 steps that are likely to reflect recombination hot spots. The mean swept radius (the distance over which LD is likely to be useful for mapping) is 13.3 Mb, confirming extensive LD in Holstein–Friesian dairy cattle, which makes such populations ideal for whole-genome association studies.

NetView: a high-definition network-visualization approach to detect fine-scale population structures from genome-wide patterns of variation.

High-throughput sequencing and single nucleotide polymorphism (SNP) genotyping can be used to infer complex population structures. Fine-scale population structure analysis tracing individual ancestry remains one of the major challenges. Based on network theory and recent advances in SNP chip technology, we investigated an unsupervised network clustering method called Super Paramagnetic Clustering (Spc). When applied to whole-genome marker data it identifies the natural divisions of groups of individuals into population clusters without use of prior ancestry information. Furthermore, we optimised an analysis pipeline called NetView, a high-definition network visualization, starting with computation of genetic distance, followed clustering using Spc and finally visualization of clusters with Cytoscape. We compared NetView against commonly used methodologies including Principal Component Analyses (PCA) and a model-based algorithm, Admixture, on whole-genome-wide SNP data derived from three previously described data sets: simulated (2.5 million SNPs, 5 populations), human (1.4 million SNPs, 11 populations) and cattle (32,653 SNPs, 19 populations). We demonstrate that individuals can be effectively allocated to their correct population whilst simultaneously revealing fine-scale structure within the populations. Analyzing the human HapMap populations, we identified unexpected genetic relatedness among individuals, and population stratification within the Indian, African and Mexican samples. In the cattle data set, we correctly assigned all individuals to their respective breeds and detected fine-scale population sub-structures reflecting different sample origins and phenotypes. The NetView pipeline is computationally extremely efficient and can be easily applied on large-scale genome-wide data sets to assign individuals to particular populations and to reproduce fine-scale population structures without prior knowledge of individual ancestry. NetView can be used on any data from which a genetic relationship/distance between individuals can be calculated.

Composite selection signals can localize the trait specific genomic regions in multi-breed populations of cattle and sheep

BackgroundDiscerning the traits evolving under neutral conditions from those traits evolving rapidly because of various selection pressures is a great challenge. We propose a new method, composite selection signals (CSS), which unifies the multiple pieces of selection evidence from the rank distribution of its diverse constituent tests. The extreme CSS scores capture highly differentiated loci and underlying common variants hauling excess haplotype homozygosity in the samples of a target population.ResultsThe data on high-density genotypes were analyzed for evidence of an association with either polledness or double muscling in various cohorts of cattle and sheep. In cattle, extreme CSS scores were found in the candidate regions on autosome BTA-1 and BTA-2, flanking the POLL locus and MSTN gene, for polledness and double muscling, respectively. In sheep, the regions with extreme scores were localized on autosome OAR-2 harbouring the MSTN gene for double muscling and on OAR-10 harbouring the RXFP2 gene for polledness. In comparison to the constituent tests, there was a partial agreement between the signals at the four candidate loci; however, they consistently identified additional genomic regions harbouring no known genes. Persuasively, our list of all the additional significant CSS regions contains genes that have been successfully implicated to secondary phenotypic diversity among several subpopulations in our data. For example, the method identified a strong selection signature for stature in cattle capturing selective sweeps harbouring UQCC-GDF5 and PLAG1-CHCHD7 gene regions on BTA-13 and BTA-14, respectively. Both gene pairs have been previously associated with height in humans, while PLAG1-CHCHD7 has also been reported for stature in cattle. In the additional analysis, CSS identified significant regions harbouring multiple genes for various traits under selection in European cattle including polledness, adaptation, metabolism, growth rate, stature, immunity, reproduction traits and some other candidate genes for dairy and beef production.ConclusionsCSS successfully localized the candidate regions in validation datasets as well as identified previously known and novel regions for various traits experiencing selection pressure. Together, the results demonstrate the utility of CSS by its improved power, reduced false positives and high-resolution of selection signals as compared to individual constituent tests.

netview p: a network visualization tool to unravel complex population structure using genome-wide SNPs.

Network‐based approaches are emerging as valuable tools for the analysis of complex genetic structure in wild and captive populations. netview p combines data quality control with the construction of population networks through mutual k‐nearest neighbours thresholds applied to genome‐wide SNPs. The program is cross‐platform compatible, open‐source and efficiently operates on data ranging from hundreds to hundreds of thousands of SNPs. The pipeline was used for the analysis of pedigree data from simulated (n = 750, SNPs = 1279) and captive silver‐lipped pearl oysters (n = 415, SNPs = 1107), wild populations of the European hake from the Atlantic and Mediterranean (n = 834, SNPs = 380) and grey wolves from North America (n = 239, SNPs = 78 255). The population networks effectively visualize large‐ and fine‐scale genetic structure within and between populations, including family‐level structure and relationships. netview p comprises a network‐based addition to other population analysis tools and provides user‐friendly access to a complex network analysis pipeline through implementation in python.

A comparative location database (CompLDB): map integration within and between species

We have adapted the Location Database (LDB) map-integration strategy of Morton et al. [Ann Hum Genet 56:223–232] (1992) as above to create an integrated map for each of several species for which fully annotated genome sequences are not yet available (sheep, cattle, pig, wallaby), using all types of partial maps for that species, including cytogenetic, linkage, somatic-cell hybrid, and radiation hybrid maps. An integrated map provides not only predictions of the kilobase location of every locus, but also predicts locations (in cM) and cytogenetic band locations for every locus. In this way a comprehensive linkage map and a comprehensive cytogenetic map are created, including all loci, irrespective of whether they have ever been linkage mapped or physically mapped, respectively. High-resolution physical maps from annotated sequenced species have also been placed alongside the integrated maps. This has created a powerful tool for comparative genomics. The LDB map-integration strategy has been extended to make use of zoo-FISH comparative information. It has also been extended to enable the creation of a “virtual” map for each species not yet sequenced by using mapping data from fully sequenced species. All of the partial maps, together with the integrated map, for each species have been placed in a database called Comparative Location Database (CompLDB), which is available for querying, browsing, or download in tabular form at http://medvet.angis.org.au/ldb/.

A Meta-Assembly of Selection Signatures in Cattle

Since domestication, significant genetic improvement has been achieved for many traits of commercial importance in cattle, including adaptation, appearance and production. In response to such intense selection pressures, the bovine genome has undergone changes at the underlying regions of functional genetic variants, which are termed “selection signatures”. This article reviews 64 recent (2009–2015) investigations testing genomic diversity for departure from neutrality in worldwide cattle populations. In particular, we constructed a meta-assembly of 16,158 selection signatures for individual breeds and their archetype groups (European, African, Zebu and composite) from 56 genome-wide scans representing 70,743 animals of 90 pure and crossbred cattle breeds. Meta-selection-scores (MSS) were computed by combining published results at every given locus, within a sliding window span. MSS were adjusted for common samples across studies and were weighted for significance thresholds across and within studies. Published selection signatures show extensive coverage across the bovine genome, however, the meta-assembly provides a consensus profile of 263 genomic regions of which 141 were unique (113 were breed-specific) and 122 were shared across cattle archetypes. The most prominent peaks of MSS represent regions under selection across multiple populations and harboured genes of known major effects (coat color, polledness and muscle hypertrophy) and genes known to influence polygenic traits (stature, adaptation, feed efficiency, immunity, behaviour, reproduction, beef and dairy production). As the first meta-assembly of selection signatures, it offers novel insights about the hotspots of selective sweeps in the bovine genome, and this method could equally be applied to other species.