Sparse models for imaging genetics

Imaging genetics is an emerging and promising technique that investigates how genetic variations affect brain development, structure, and function. Many imaging genetic studies involve disorder-related neuroimaging phenotypes and molecular genetics data, such as single nucleotide polymorphism data. However, this class of studies is challenging due to the relatively small number of subjects but extremely high-dimensionality of both imaging and genetic data. In this chapter, we introduce a suite of sparse methods-that can produce interpretable models and are robust to overfitting-for imaging genetics. Moreover, we can incorporate various biological prior knowledge-such as linkage disequilibrium information-into the analysis. However, due to the nonsmooth and highly complex regularizers, the applications of sparse models to large-scale problems remain challenging. Thus we introduce a suite of novel optimization techniques, that is, sparse screening, which can boost the efficiency of many sparse models on large-scale data sets by several orders of magnitude.