Benefit of Bayesian Clustering of Longitudinal Data: Study of Cognitive Decline for Precision Medicine

Precision medicine is an emerging field that aims to improve disease prevention and treatment by accounting for individual variations in genes, lifestyle and environment. Extensive datasets are used to uncover subpopulations with different biological characteristics and susceptibility to the disease or response to treatment. This finer characterisation is ultimately directed towards the identification of clinical biomarkers, the prediction of the risk of disease and its progression, or the estimation of treatment effects at the individual level, based on the subject’s shared characteristics with one of the identified subpopulations. This chapter focuses on the study of cognitive decline and brain imaging for precision medicine, and is based on data from the North American ADNI cohort. We aim to identify clusters of subjects with different susceptibility to cognitive decline, based on repeated cognitive scores and baseline MRI volumetric data. However, the clustering analysis of such an observational dataset raises a number of challenges. We describe in this chapter how Bayesian modelling can be used to overcome them successfully.