This proposal describes the development of statistical genetic methodology to increase the accuracy of genomic prediction for complex phenotypic traits by optimal use of sequence information. This new prediction methodology will approximate as closely as possible total genetic variation, i.e., between genotype variation, by the variance in genomic estimated breeding values (GEBVs), where the latter GEBVs will be based on sequence information.
These new prediction methods will depend on the identification of the causal SNPs/genomic regions in contrast to existing prediction methods that depended on long-ranging within-family linkage disequilibrium (LD) between SNPs and QTLs. With statistical and bioinformatics approaches, we will first identify the potential contributions of different types of genetic variation to total genetic variation, and then develop optimal differential weighting schemes for the various genetic effects in genomic prediction models. We will also investigate design questions for genomic prediction: what are the main factors determining accuracy of prediction and its persistence across generations, populations and conditions? Furthermore, we will pay attention to models allowing training on multiple traits/environments to predict a broad breeding target. Ultimately, our prediction models should approach or exceed 85% accuracy with sufficient persistence for the accuracy of a target trait in pure line, multiple line and cross bred prediction.
The main objectives of the project are to
- Evaluate various methods of genomic prediction within and across lines
- Investigate ways to improve power for GWAS in populations with strong family structure and long range LD.
- Develop optimal genomic prediction strategies for new and expensive traits where small training populations will be common.
- Derive general rules for optimal phenotyping and genotyping, including composition of training set, by analytical and simulation approaches.
- Develop methodology for genomic prediction within and across breeds
- Develop methods for genomic prediction that exploit local LD structure and information on for candidate genes.
- Prof. Fred van Eeuwijk, Biometris, Wageningen UR
- Prof. Roel Veerkamp, Animal Breeding and Genomics Center, Wageningen UR
- Prof. Jeanine Houwing-Duistermaat, Leiden University Medical Center
3 PhD candidates with relevant background for research on genomic prediction methods (quantitative genetics, statistics, and/or genomics). The PhD candidates will work on three different species: dairy cows, pigs and poultry.
Further information: please contact one of the main supervisors