Thèse: Incorporating known functional annotations into Bayesian genomic prediction models
Genomic selection has been successfully implemented in many livestock breeding programs in the last decade, and one recently proposed means for improvement is the use of underlying biology as an additional source of information to guide predictions for complex traits. The aim of this thesis project is to develop and validate genomic prediction models able to weight SNPs to incorporate information extracted from functional annotation maps obtained at different molecular levels on top of the commonly used phenotypic and genotypic data. Subsequently, the extent to which this may increase prediction accuracy will be investigated. The candidate will develop novel cutting-edge statistical models to integrate new functional information into existing genomic predictions methods and validated in commercial populations. In particular, the thesis will focus on the development of a computationally tractable empirical Bayesian genomic prediction model that incorporates functional annotation into prior distributions to weight the likelihood that a given genetic variant is functional or has predicted functional impact on a phenotype. This thesis project thus represents the development of an innovative and impactful approach, making use of statistically robust approaches, to potentially have a major influence on breeding practices.
The thesis project is anchored in data that will be collected in the multi-actor GENE-SWitCH (The regulatory GENomE of SWine and CHicken: functional annotation during development) EU Horizon 2020 Research and Innovation project. GENE-SWitCH supports the efforts of the ongoing Functional Annotation of ANimal Genomes (FAANG) consortium, a worldwide consortium of 350+ contributors with the global goal of establishing reference functional maps of domesticated animal genomes. GENE-SWitCH will produce new genomic information to enable the characterization of genetic and epigenetic determinants of complex traits in the two monogastric species (chicken and pigs) that are the primary sources of meat worldwide. In particular, by producing comprehensive reference functional annotation maps at several molecular levels (RNA-seq, ATAC-seq, ChIP-seq, Hi-C) in tissues of relevance to sustainable production at different time points (from embryo/fetus to adult life), the GENE-SWitCH project aims to integrate functional information into genomic selection schemes to improve their effectiveness in the pig and poultry sectors. The large amount of biological information that will be generated in the GENE-SWitCH project will thus provide an excellent opportunity to empirically investigate the potential benefit of using functional annotation in genomic prediction.