ABSTRACT
Mixture models are intensively used in genetics and genomics either for identifying latent structures or for modeling densities. According to the type of mixture component and the nature of the hypothesis about latent structures, mixture models may be relevant in numerous different frameworks. In this chapter, the use of mixture models in genetic and genomics is presented by increasing complexity of the latent structure. Section 18.2 considers applications with independent latent variable structures to genome and transcriptome analysis. Section 18.3 illustrates the use of hidden Markov models (HMMs) in genomics, presenting a variety of problems with their associated translation in terms of emission distributions and hidden states. Finally, Section 18.4 introduces more complex dependency structures used in genomics such as the hidden Markov random field and stochastic block model with their associated parameter estimation difficulties.
