TITAN: A Toolbox for Information-Theoretic Analysis of Molecular Networks

Molecular, cellular, structural, and functional networks in the brain each represent different levels of organization and complexity. These networks are interconnected, and their interactions underlie the brains abilities to process information, regulate bodily functions, and mediate behavior. While understanding the emergence of higher-level networks (structural and functional) from the interactions of lower-level (molecular and cellular) ones is crucial, a universal analytical method applicable to all brain network types has been lacking. Here we introduce an open-source toolbox (NETSCOPE) that identifies weighted network architectures using mutual information (MI) and variation of information (VI). We demonstrate the accuracy of the resulting networks using synthetic data and by recreating five molecular networks in S. cerevisiae. We finally deploy NETSCOPE to identify cell type specific transcriptional networks and reconstruct the brain-wide neural networks that encode touch in the mouse brain. The code is made available in Python, Google Colab, Jupyter Notebook, MATLAB, and Octave. Beyond its core functionality, this network discovery method holds potential for broader applications, including the development of bioinspired sparse artificial networks.