Seminar on Computational Learning and Adaptation

A Framework for Representing, Evaluating,
and Revising Complex Biological Models

Stephen Racunas and Nigam Shah
The Huck Institutes of the Life Sciences
Pennsylvania State University
www.hybrow.org

In the current content-rich environment, building and evaluatingbiological models requires scientists to collect, evaluate, and integrate information of many disparate kinds. A unified frameworkthat represents and relates diverse biological data and processeswould provide clear benefits to the scientific enterprise. In response, we have developed the HyBrow (Hypothesis Browser) environment. This system incorporates a formalism for stating biological models and phenomena at different levels of detail and linking them to different biological information sources, graphical tools for visualizing and editing these models, methods for checking the consistency of model components against accepted phenomena, and techniques for proposing revisions that would remedy detected problems. We intend Hybrow not as an automated discovery system, but rather as an interactive tool to support biologists' development of explanatory models. We demonstrate our prototype environment by using it to evaluate alternative hypotheses related to the galactose gene network in Saccharomyces cerevisiae.

This talk presents work done jointly with I. Albert and N. Fedoroff.

Note: Although this work does not focus on computational learning, it deals with issues that arise in knowledge-rich approaches to scientific modeling and discovery.

Return to the seminar schedule