A Selective Overview of Fusion Penalized Learning in Latent Subgroup Analysis for Precision Medicine

Modern technology has facilitated the collection of large-scale data from electronic medical records, health insurance databases and other platforms. Big data along with rapid growth in computational power are creating unprecedented opportunities in precision medicine that is at the forefront of medical research. Understanding the disease heterogeneity is essential to the development of precision medicine that aims to tailor treatments to subgroups of patients with similar characteristics. One major challenge of achieving this goal is that the heterogeneity can be driven by latent factors which are unobservable in the datasets. This article gives a review on a pairwise fusion penalized learning method for latent subgroup analysis. This approach has received increasing attention in recent years. The problem of subgroup investigation can be formulated into a heterogeneous regression model that allows the parameters to be subject-dependent with unknown grouping information. We then apply the fusion penalized method that can automatically divide the patients into different groups as well as estimating the parameters simultaneously. This machine learning approach is flexible, can be adapted into different regression models and data settings, and requires mild assumptions. We will discuss in detail the computational algorithms for this method, and its applications to longitudinal data with different types of responses such as continuous, binary and count data. We will also illustrate the performance of this method through simulation studies and a biomedical data application.