Using Genetic Relationships to Improve the Design and Analysis of Animal Science Studies

It is well established that if identifiable blocking factors account for a substantial proportion of the variability for key traits of experimental interest, then randomly assigning animals to treatments within blocks should increase statistical power. In fact, block designs also generally lead to inferences that are more robust and reproducible provided that the blocks chosen for the study are widely variable and representative of the intended target population. For moderately to highly heritable traits, blocking on families should be effective and relatively straightforward to conduct for litter-bearing species such as pigs compared to, say, cattle for example. It seems then that genetic or genomic relationships between animals should be taken into consideration when blocking for treatments in dairy or beef cattle studies. We statistically assess the benefits of blocking in traditional arrangements of large half sib or full sib families as functions of heritabilities, effect sizes, and number of families. However, recognizing that population structures may be far more complex than large sib families for cattle research, we also assess the benefits of blocking based on general pedigree and/or genomic relationship matrices as well. This blocking or clustering can be based on principal component analyses, for example, which is routinely used in quantitative genomics to identify population structure. As with traditional blocking factors, genetic effects can be readily modeled as random effects within a mixed effects model. Power analyses based on mixed effects modeling is reviewed and extended to account for more general population genetic structures compared to classical block designs. We also discuss how degrees of freedom (i.e., true biological replication) for such tests might be more appropriately inferred, particularly when genetic or family effects are partially confounded with or nested within treatments. The implications for multi-pen and multi-herd studies when the experimental unit is pen or herd are discussed in the context of the degree of genetic connectedness between pens or herds. The implications of genotype by environment and/or genotype by treatment interaction on the design of animal studies are addressed as well. Properly accounting for genetic effects, particularly for moderately to highly heritable traits, should improve research reproducibility and facilitate a better assessment of the potential for precision management of livestock based on their genotypes and/or pedigrees.