On Metabolic Pathway Reconstruction from Gene Expression Data
Cedric Gondro1, Brian P. Kinghorn2
1genetics@sigex.com.br, University of New England; 2bkinghor@une.edu.au, University of New England
This work aims to infer metabolic pathways and other biological processes from data generated in microarray experiments. This is a challenging task, due to the many candidate components and the complexity of the underlying mechanisms. However, methods can be developed and tested on the basis of gene and protein expressions simulated using relatively simple biological models. One approach is to use kinetic simulations of biological molecules, based on empirical biological knowledge of the system under test. This may be a useful approach for simulating gene expression data in complex situations for which an analytical approach may become intractable. It also allows a high degree of ‘biological analogy’, and converting simulation output into gene expression data formats is a trivial exercise. Expression datasets generated will be analysed using evolutionary algorithms, such as genetic programming, in an attempt to identify potential methods for inferring biological mechanisms. Candidate solutions will be parameterised models of the biological system. Many thousands of these will be generated and tested as part of an evolutionary algorithm, and this means that the computationally slower kinetic models are likely to be replaced by dose-response type models during this phase. A preliminary test has shown correct reconstruction of lac operon model parameters derived from simulated expression data collected following a perturbation in the level of lactose. For full model construction, an evolutionary algorithm would draw on a set of generic biological components and their relations to reconstruct the candidate “biological mechanism” solutions. Such components would include substrates, enzymes, hormones and receptors. Objective functions for this algorithm might involve goodness of fit to expression results, physiological measurements, and prior knowledge about metabolic pathways.