Core engine for genomic prediction

Genomic prediction is the state-of-the-art method used for genetic evaluations in modern breeding schemes. Embraced by all Breed4Food partners, genomic prediction evaluations are routinely performed using available phenotypes, pedigree, and genotypes, to provide the latest and most accurate breeding values for selection candidates. Given this crucial role of genomic prediction, and with the increasing amounts of phenotypes and genotypes, this workpackage (WP1) “Core engine for genomic prediction” aims to deliver state-of-the-art fast and reliable statistical methods, that are implemented in MiXBLUP, that is a stable software for very large genetic evaluations.

All breeding companies use software, pipelines and models as their core engine for routine genetic evaluations in their breeding schemes. These genetic evaluations need to provide  accurate and unbiased breeding values for selection candidates, at a weekly or even daily basis. The state-of-the-art method for these routine genetic evaluations is currently the so-called single-step genomic Best Linear Unbiased Prediction, that analyzes simultaneously phenotypic, pedigree and genomic evaluations. Given its crucial role in animal breeding companies, research to innovate this core engine has been massive both in terms of software and model developments in the past 15 years. The Breed4Food consortium has been a global leading centre in this area. Research and development within the Breed4Food consortium were consolidated in MiXBLUP (, a versatile software for large genomic evaluations. While major developments have been successfully implemented, further innovations are continuously needed, to accommodate faster evaluations able to run on the newest infrastructure and to meet the current deadlines, especially with the continuously increasing phenotypic and genomic datasets. Other developments include single-step genomic evaluations for multiple breeds and crossbreds, and computation of the reliability of the estimated genomic breeding values.