WITHDRAWN: Considering Developmental Change in Gxi Research and Links to Replication

Genetic associations and gene-environment interactions (GxE) are developmentally contingent such that some are expected to manifest during specific developmental periods and not others. Integrating development into molecular genetic research in psychology is one of the several next-steps for a post-GWAS research era and holds promise for identifying critical periods for genetic and environmental influences important for behavioral interventions. Considering development in GxE and gene-by-intervention interaction (GxI) research will be informative for understanding how and when these transactions manifest (and when they do not). Further, greater integration of development may also help clarify some of the replication issues associated with GxE research. Historically speaking, addressing replication issues in GxE research has focused on producing sets of direct replications by homogenizing analytic models and variables across samples. Even direct replications that use identical measures and models may fail to reproduce findings, however, if the samples have meaningful heterogeneity, such as different age distributions or focal developmental periods. Given that many GxE and GxI studies rely on relatively small samples (N < 1,000) and circumscribed age ranges, developmental considerations may be particularly important for identifying reliable associations in these research areas.

The focus of the current talk is on describing how better integrating development into GxI research can help a) identify sensitive periods for intervention effectiveness, b) determine developmental stages when genetic associations are more probable, and c) model developmental change in GxI. Recent research using data from the genetic subsample of the PROSPER project (gPROSPER) will be presented that use time-varying effects models (TVEM) that examine developmentally contingent genetic, intervention, and GxI associations. TVEM is an extension of ordinary-least squares regression that permits bivariate associations to vary as a function of time, such as participant age, and is ideal for evaluating developmental hypotheses. Specifically, research on the time-varying association between the GABRA2 gene, the PROSPER-delivered interventions, and GxI on adolescent delinquency will be presented as an example of developmentally contingent genetic and GxI associations. In addition, replication research of one of the first published GxI studies will be discussed in conjunction with analyses that demonstrate how greater integration of development using TVEM can help clarify replication failures.