Many longitudinal neuroimaging studies collect genetic and recurrent imaging data to track individual changes in brain structure and function over time. Several neurodevelopmental and neurodegenerative disorders can arise due to abnormal development of the brain, which may be caused by the additive and/or interactive effects of various risk genes and environmental risk factors, each contributing small individual effects. Recurrent neuroimaging can reveal the genetic pathways and causal genes associated with the specific brain changes that underlie such neurodegenerative disorders. We developed the Bayesian longitudinal low-rank regression (L2R2) model to characterize the temporal change patterns of high-dimensional neuroimaging responses and analyze the association between high-dimensional longitudinal neuroimaging responses and high-dimensional genomic covariates and demographic variables. We also used the correlation structure of the high-dimensional neuroimaging responses and genetic markers to increase the detection power of important associations. The L2R2 model has increased power to detect important pairs of imaging and genomic variables that cannot be detected by traditional pair analysis. The model can identiy age-dependent genetic factors, which help characterize the genetic effects on brain morphology over time and reveal time-varying genetic factors underlying age-related disorders.
Lu ZH, Zakaria K, Ibrahim JG, Wang Y, Zhu HT. Bayesian longitudinal low-rank regression models for imaging genetic data from longitudinal studies. NeuroImage 149(1):305-322, 2017. doi:10.1016/j.neuroimage.2017.01.052. PMID: 28143775.