Recent advancements in medical and other fields of scientific research have allowed scientists to collect data of unprecedented size and complexity. A common statistical problem in these applications is to model a response variable of interest as a function of a small subset of a large number of covariates (features). The problem becomes even more complex when a population under study is made up of hidden sub-populations and the relationship between the response variable and covariates varies across sub-populations. Mixture-of-experts (MOE) models provide a flexible statistical tool for studying such relationships. In this talk we discuss some new developments on estimation and feature selection methods in MOE models with diverging number of parameters.