Multivariate functional data, where continuous observations are sampled from a vector of random functions, are emerging in a wide range of scientific applications. A central problem in multivariate functional data analysis is to investigate conditional dependence among the functions. This can be formulated as graphical modeling of multivariate functional data. Most existing graphical models assume the data are sampled from random variables that follow multivariate Gaussian or copula Gaussian distributions, and the relations among the functions on the nodes are linear. In this talk, we discuss some ongoing research projects that extend graphical modeling from random variables to random functions, and from Gaussian models to nonparametric models. We discuss some potential applications in brain connectivity analysis.