Automated Inference of Symbolic Models for Gene Regulatory Networks

Motivation: Multiple models exist to describe the topology and the dynamical behavior of gene regulatory networks (GRNs). Among them, symbolic models (such as systems of ordinary differential equations) have the advantage of providing intuitive insights into the inner workings of a network so that general laws can be derived. Obtaining such models from time series data has proven to be a difficult task because biologists must rely on existing first principles. In this project, I propose to enhance the use of grammatical evolution as an alternative to existing work on automatic extraction of symbolic models for GRNs from temporal data. Results: I present GERNI (Grammatical Evolution Regulatory Network Inferrer). This tool uses grammatical evolution to infer a system of differential equations that describes a given time series. My results indicate that GERNI is able to infer the structure of certain classes of systems and that it does so within a small number of generations. Nonetheless, further work is necessary to ensure that the program is able to deal with more complex types of data. Availability: All computer code will be hosted at http://www.cs.colostate.edu/~andrescj Contact: [email protected] Supplementary information: N/A.