LUCIDus: An R Package For Implementing Latent Unknown Clustering By Integrating Multi-omics Data (LUCID) With Phenotypic Traits

Many studies are leveraging current technologies to obtain multiple omics measurements on the same individuals. These measurements are usually cross-sectional, and methods developed and commonly used focus on omic integration at a single time point. More unique, and a growing area of interest, are studies that leverage biology or the temporal sequence of measurements to relate long-term exposures or germline genetics to intermediate measures capturing transitional processes that ultimately result in an outcome. In this context, we have previously introduced an integrative model named Latent Unknown Clustering by Integrating multi-omics Data (LUCID) aiming to distinguish unique effects of environmental exposures or germline genetics and informative omic effects while jointly estimating subgroups of individuals relevant to the outcome of interest. This multiple omics analysis consists of early integration (concatenation of omic layers to estimate common subgroups); intermediate integration (omic-layer-specific estimation of subgroups that are all related to the outcome); and late integration (omic-layer-specific estimation of subgroups that are then interrelated by a priori structures). In this article, we introduce LUCIDus version 3, an R package to implement the LUCID model. We review the statistical background of the model and introduce the workflow of LUCIDus, including model fitting, model selection, interpretation, inference, and prediction. Throughout, we use a realistic but simulated dataset based on an ongoing study, the Human Early Life Exposome Study (HELIX), to illustrate the workflow.