A combined coalescence forward in time simulator software for pedigreed populations undergoing selection for complex traits

Abstract Text: The use of marker information in animal breeding has recently been an active area of research and has been incorporated in selection decisions and as a tool to control inbreeding across a variety of species. There is yet still much to be learned on the optimal way to use marker information to select animals and manage the genome of a population that is undergoing selection for complex traits that have a traditional quantitative basis (i.e. yield) and/or fitness basis (i.e. number of progeny). We have developed a combined coalescence and forward-in-time simulator for complex traits and populations. The simulator is carried out in two stages. In the first stage whole-genome SNP data is read in ms format and is utilized to generate founder individuals and associated SNP marker panels ranging in size from thousands to millions of SNP. During this stage a wide variety of trait architectures can be generated with additive and dominance effects for both a traditional quantitative trait and fitness along with genomic covariance among traits. The second stage generates new individuals across generations based on a variety of selection scenarios. The selection stage can be performed using a wide variety of relationship matrices including pedigree, independent markers, haplotypes, or run of homozygosity based haplotypes. Relationship matrices and their associated inverse are generated using computationally efficient algorithms based on updating matrices from previous generations. Complex population structures can be generated that allow for a differential contribution of gametes to the next generation as well as mating constraints. To demonstrate the program, we present a small application that mimics a dairy cattle and swine population in order to describe some of the metrics that are generated. Scenarios were generated based on a 12,000 SNP marker panel spread across 3 chromosomes and a population size of 650 animals (sires=50; dams=600) per generation. A scenario with selection on a quantitative trait occurring for 5 generations and breeding values estimated from pedigree or independent SNP had a running time for the dairy cattle scenario of 4.85 and 5.82 minutes, respectively. Geno-Driver allows for a wide range of selection strategies to be evaluated in the presence of a fitness trait and is available at https://github.com/jeremyhoward/GenoDriver.

Keywords:

genetic simulation, quantitative traits, genomic selection