"Modeling pleiotropic genetic systems in mice"
Variation of complex traits in a genetically randomized population of laboratory mice (GPRLM) can provide observations that shed light on the system’s structure. In biological systems, experimental crosses provide unique features that admit causal inference. Naturally occurring genetic variation can be distributed in a combinatorial fashion among a set of cross progeny. The transmission of genetic variants through the process of meiosis serves as a natural randomization mechanism. Thus genetic crosses share properties of statistical designed experiments that allow us to infer causation. This is consistent with the intuitive notion that causation flows from genes to phenotypes. We have applied methods of network inference to GRPLM to investigate a variety of phenotypes. A typical study involves the measurement of a number of correlated phenotypes and genotyping on the same set of mice. The problems to be addressed are three. First, how to identify the genetic loci involved in the system? Second, how to infer the network of causal relationships among genes and phenotypes? Third, what can we learn from the network? We illustrate this concept with an application to analyses of three complex trait systems. We are able to identify genetic loci that are causal for variation in multiple phenotypes. Moreover we are able to infer causal networks that provide insight into the pleiotropic nature of the genetic effects.