Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Animal Genomics and Improvement Laboratory » Research » Research Project #433412

Research Project: Improving Dairy Animals by Increasing Accuracy of Genomic Prediction, Evaluating New Traits, and Redefining Selection Goals

Location: Animal Genomics and Improvement Laboratory

Project Number: 8042-31000-002-00-D
Project Type: In-House Appropriated

Start Date: Jul 24, 2017
End Date: Jul 23, 2022

Objective 1. Expand genomic data used in prediction by selecting new variants that more precisely track the true gene mutations that cause phenotypic differences. Objective 2. Evaluate new traits that can all be predicted at birth from the same inexpensive DNA sample. Objective 3. Improve efficiency of genomic prediction and computation by developing faster algorithms, testing new adjustments and models, and accounting for genomic pre-selection in evaluation.

Obj. 1: Variant selection strategies will be tested with 1000 Bull Genomes data. Two-stage imputation will be used; imputation accuracy will be compared by simulation. Local sequence data will be generated for families with new fertility defects or other health conditions and bulls homozygous for less frequent haplotypes. Animals will be selected for sequencing with an algorithm maximizing coverage of rare haplotypes and minimizing resequencing of common haplotypes. Previous data will be realigned to a new reference map. Candidate variants will be reselected using improved annotation, better bioinformatics, and information from discoveries across species. Lists of candidate variants with the largest effects will be supplied for array design. Best strategies to include gene-edited animals in breeding programs, their potential value, and confirmation of phenotypic effects of gene edits will be determined. Simulation will reveal optimum strategies for combining favorable haplotypes. Obj. 2: Genetic evaluations will be developed for traits already measured but with low heritability or moderate economic value. Economic values and reliability for new traits will be estimated; options for choosing the most profitable animals to phenotype and genotype will be explored. Data editing and analysis methods will be developed for new data. Computer simulation will be used to determine the best combination of direct and indirect phenotypes for genetic improvement. Relative economic values will be calculated for selection indexes; index sensitivity will be determined based on forecast economic value. Selection index methodology will be used to study effect on annual rates of genetic gain from adding recessives to the index. Incidence, correlations, and effects of more traits will be documented. Constant monitoring of input data will ensure continued high-quality evaluations. Obj. 3: Algorithms will be developed to improve aligning sequence segments to a reference genome while simultaneously calling variants. Genomic models will be designed to include more informative priors. Tests will compare predictive ability for future data within or across breed. Multibreed marker effects will be estimated as correlated traits. Potential biases from genomic pre-selection will be monitored using differences across time in percentages of genotyped mates or daughters. Use of single-step models to correct bias will be explored using recent algorithms to approximate the inverse of genomic relationships and model marker effects directly. Genomic evaluations of crossbred animals will be developed by weighting marker effects from each breed by genomic breed composition. Prediction of nonadditive effects and recombination loss will be continued. Genomic future inbreeding will be improved by computing average genomic relationship to a more recent group of potential mates instead of to breed reference population. Test-day models will be considered when appropriate. Adjustments will be tested using truncated data to predict more recent data. Multitrait processing will be used to obtain greater benefits from new traits without losing information from previous correlated traits.