gen3DNet: An R Package for Generating 3D Network Models

Abstract

MotivationNetworks are ubiquitous to organize and represent data in the real world such as the biomedical networks, social networks, etc. Due to the attractive property of networks which can explicitly capture the rich relationships among data samples, there is an increasing demand to transform linkage-free data to graph-structure data for efficient downstream analyses. Existing works typically focus on representing a single data object like gene expression profile in a homogeneous 2D network, which fail to deal with situations where two different data objects are involved to create a 3D heterogeneous network. ResultsIn this paper, we introduce an R package, gen3DNet (a generic version of the iPhDNet), for generating 3D network models from two correlated objects with shared common factors. Specifically, gen3DNet builds the relationships between samples and shared factors using the non-negative matrix factorization, where three clustering techniques are evaluated to determine the number of functional modules. In addition, it builds the relationships between samples from the two distinct data objects based on a partial least squares regression model. Usage of the package is illustrated through a real-world application. GitHub URL Source code is available at https://github.com/mollahlab/gen3DNet