Fig. 1: Gene function, in vitro and in vivo phenotypes.

(1) Effects of experimental perturbations on an in vitro cellular phenotype measured under experimental conditions are used as proxies for the effects of genetic LoF variants on a difficult-to-measure in vivo cellular phenotype. In vitro cellular models are generally within a narrow biological context with specific cell types and physiological states. Cellular models may be in non-human model organisms (e.g., mice), which will impact proxy effectiveness. (2) The corresponding in vivo cellular phenotypes can have downstream effects. If we assume that higher in vivo intra-cellular lipid content has a causal effect on increased serum LDL-cholesterol, it follows that genes influencing intra-cellular lipid content also influence serum LDL-cholesterol. For example, if loss-of-function in the LDL receptor gene (LDLR) increases intra-cellular lipid content, then loss-of-function in LDLR should also increase serum LDL-cholesterol. In vitro functional screens experimentally perturb gene function to identify genes where reduced (or increased) function affects an in vitro cellular phenotype. In vivo population studies (e.g., in UK Biobank) estimate the effects of germline variants lowering gene function on measured phenotypes in a wide biological context (all cells, across the life-course). GPAT combines these data at the gene level to evaluate the transportability of effect estimates from in vitro cellular models to in vivo measured phenotypes. GPAT estimates capture similarities between the in vitro model and in vivo cellular phenotypes (1) and causal relationships between in vivo phenotypes across a wide biological context (2).