Primary methodologies/approaches/strategies used to accomplish the goals

We are actively developing a three-step network inference procedure wherein biclusters (i.e. subsets of genes with similar expression profiles over a subset of experimental conditions) are identified using an algorithm termed cMonkey. This is an iterative procedure guided by a wide range of computationally- and experimentally-derived functional relationships among genes, including conserved phylogenetic profiles, operon organization and metabolic networks, protein-protein and protein-DNA interactions, and genome wide measurements of mRNA and protein level changes. In each iteration, upstream sequences of genes in each bicluster are simultaneously evaluated for co-occurrence of statistically significantly conserved cis-regulatory elements.

Subsequently, the Inferelator screens, for each bicluster, all of the putative transcription regulators and environmental factors to identify the best set of descriptors that together sufficiently predict the dynamic behavior of the genes in that bicluster. This step yields a raw regulatory network that may be perturbed or evaluated for robustness in the third step by algorithms such as the Evolvolator. Herein, the dynamics of the network are interrogated in an attempt to recapitulate what was previously observed or to predict novel dynamics which may be verified by further experiments.

Future directions of this project.

Future research in this project is aimed at improving and automating this network inference procedure. One of the goals is to add features that will enable intelligent curation by experts exploring previous outputs.

How does the project address the issues of predictive, preventive or personalized medicine?

The methods for deciphering the predictive models delineated here will have direct impact on designing similar procedures for modeling complex disease models. Such a quantitative predictive model could enable early disease diagnosis, and help design preventive and personalized medicine.

Group members involved with Project

• Nitin S. Baliga (PI)
• Richard Bonneau (Co-PI)
• David Reiss

Internal ISB groups/people involved with project.

• Leroy Hood
• Vesteinn Thorsson