Information theory, multivariate dependence, and genetic network inference

We define the concept of dependence among multiple variables using maximum entropy techniques and introduce a graphical notation to denote the dependencies. Direct inference of information theoretic quantities from data uncovers dependencies even in undersampled regimes when the joint probability distribution cannot be reliably estimated. The method is tested on synthetic data. We anticipate it to be useful for inference of genetic circuits and other biological signaling networks. 1 Two problems One of the most active fields in quantitative biology is the inference of biological interaction networks (e. g., protein or genetic regulatory networks) from high throughput data such as expression microarrays [1]1 . In these problems, one measures (simultaneous or serial) values of expressions of genes under different conditions and treats them as samples from a joint probability distribution (PD). The goal is to infer the genetic network based on statistical dependencies in this PD. This in...

• Type: Journal
• Year: 2004
• Where: CORR
• Authors: Ilya Nemenman
• Keywords: CORR 2004, Education, Interaction Networks, Joint Probability Distribution, Statistical Dependence