[go to GA/PLS]

The GA/PLS Multiple Metabolite TIPS^© Process¹

GA/PLS was developed to evaluate the relative importance of each gene with respect to a cellular function. We developed GA/PLS to integrate metabolic and gene expression profiles for a hepatocellular system. But, one can apply genetic algorithm and partial least square analysis to identify important genes relevant to any cellular function. We identified genes whose expression levels quantitatively predict a metabolic function and that play a part in regulating a hepatocellular function, then reconstructed their role in the metabolic network.

¹Li, Z., Chan, C., "Integrating Gene Expression and Metabolic Profiles", 2004, Journal Biological Chemistry, 279:27124-27137. [abstract]

[go to ICA/BN]

The ICA/BN 1 or 2 Metabolite TIPS^© Process²

We apply a mathematical framework that first integrates genetic algorithm (GA) and partial least squares (PLS) analyses to identify the genes relevant to cytotoxicity, as measured by lactate dehydrogenase (LDH) release, but these genes may be involved in many independent pathways. Therefore, the framework then applies constrained independent component analysis (CICA) to identify an independent pathway involved in LDH release.

ICA is a statistical method that has been applied to reveal "hidden factors" underlying sets of signal measurements. The expression levels of the genes are the recorded signals which are affected by underlying regulatory pathways.

Finally the connections between these genes selected by CICA are reconstructed using Bayesian Network (BN) analysis to infer how the genes interact with each other in the independent pathways. The reconstructed network illustrates how the genes interact under the given environmental conditions to regulate LDH release.

The advantage of the Bayesian framework over other data-driven methods is the ability of the Bayesian approach to perform cause-and-effect analysis, thus providing a basis of identifying causal relationships. This is accomplished without a priori detailed knowledge or assumptions of the biological system and the governing equations, but rather is based on the concept of conditional probability. If P(A,B|C)=P(A|C).P(B|C), then we say that A and B are conditionally independent to each other given C. Using this conditional independency test, BN tries to infer the causal relationship within a network.

²Li, Z., Chan, C. "Inferring pathways and networks with a Bayesian framework," 2004, Faseb J. 03-0475fje. [abstract]

More info on BN Learning can be found as a free for academic use downloadable program from LibB for Windows/Linux developed by researchers in Computational Molecular Biology at The Selim and Rachel Benin School of Computer Science and Engineering, The Hebrew University of Jerusalem [visit site]

[go to GSEA]

The GSEA/MBPLS Application³

To incorporate gene functional information into the analysis, we deveoped a hierarchical framework consisting of three stages to identify the processes and genes that regulate lypotoxicity. First, discriminant analysis identified that fatty acid oxidation and intracellular triglyceride accumulation were the most relevant in differentiating the cytotoxic phenotype. Second, gene set enrichment analysis (GSEA) was applied to the cDNA microarray data to identify the transcriptionally altered pathways and processes. Finally, the genes and gene sets that regulate the metabolic responses identified in step 1 were identified by integrating the expression of the enriched gene sets and the metabolic profiles with a multi-block partial least squares (MBPLS) regression model.

³Li, Z., Srivastava, S., Mittal, S., Norton, P., Resau, J., Chan, C., "A Hierarchical Approach to Identify Pathways that Confer Cytotoxicity in HepG2 Cells from Metabolic and Gene Expression Profiles," 2007, BMC Systems Biology 1:21. [abstract]

[go to TIPS^© All Stages]