Fig. 1: Study framework.
Analytical framework to study longitudinal genetic effects on phenotypes subject to age-related decline. The structural causal model is shown at the top, where the phenotype P at time point t (P0 = baseline, P1 = follow-up) varies as a function of time-invariant (baseline) genetics (G0), time-varying (longitudinal) genetics (GE) and the environment (E). Simulations are used (middle panel) to assess the suitability of three definitions of change (Δ) for longitudinal genome-wide analyses (Δ ~ γ ⋅ G), including absolute change (ΔDIFF, i.e., the absolute difference between the baseline phenotype, P0, and the follow-up phenotype, P1), relative change (ΔLOG, i.e., the difference between \(\log ({P}_{0})\) and \(\log ({P}_{1})\)) and conditional change (ΔRES, i.e., the difference between the observed P1 and the predicted \(\hat{{P}_{1}}\) phenotype). Genome-wide tests and downstream analyses (bottom panel) are performed on composite scores of cross-sectional (i.e., time-invariant) and longitudinal (i.e., time-varying) indexes of cognitive and physical aging.
