The generalized availability and decreased cost of high-throughput genotyping technologies have accelerated the development of genomic evaluations (GE) for many livestock species. The GE methods share a common objective: to precisly estimate an estimated breeding value from the effects of groupe of SNP (single nucleotide polymorphisms). To do this, Bayesian prediction models were rapidely adopted for their capacity  to simultaneously evaluate in a flexible manner the effects of SNP and also because they can incorporate a priori biological information. In parallel, several international actions are working on characterizing in different species, the functionnal intermediate processes (gene expression, methylation, chromatin accessiblity, ...) in a variety of tissues or developmental stages. The annotations built from these rich functionnal data represent heterogenous and partially overlapping information, that cannot be integrated in the Bayesian models available. 

As part of the H2020 GENE-SWitCH project, scientists from GABI developed two new Bayesian models for Genetic Evaluation (BayesRCπ and BayesRC+) to improve associations between genotypes and quantitative phenotypes while managing complex annotations. These models, published in the Journal BMC Bioinformatics and available in the open software and open BayesRCO (BayesRC for Overlapping annotations; https://github.com/fmollandin/BayesRCO), are based on two hypotheses of different overlapping biological information (incertitude vs increased confidence). The models presented have revealed to be promising for their predictive performance and interpretability on simulated and real data in pigs.
Contact :

• Dr. Andrea Rau (andrea.rau@inrae.fr) or Dr. Pascal Croiseau (pascal.croiseau@inrae.fr)