Systems engineering approaches to computational biology

Date of Completion

January 2002


Biology, Molecular|Health Sciences, Pharmacology|Engineering, Biomedical|Engineering, Chemical|Engineering, System Science




Systems engineering approaches to computational biology are developed. Different research issues in two projects, namely rational drug design and gene expression profiling from microarray experiments are addressed. ^ A concept from optimization theory, specifically mathematical programming is implemented for designing drugs with desired properties. The mathematical programming formulation is solved to obtain the optimal lead candidates, in the sense that they exhibit both high selectivity and activity while ensuring low toxicity. Both linear and non-linear models for quantitative structure activity relationships (QSAR) have been developed for use in the proposed approach. The proposed mathematical programming strategy has been demonstrated for a class of non-classical antifolates for pneumocistis carinii and toxoplasma gondii dihydrofolate reductase. The technique proposed is general and can be applied to other structure based drug design.^ Similar mathematical programming approaches are developed and implemented for gene expression profiling and tissue classification from microarray experiments. The inherent disadvantages of existing approaches and the various research issues are addressed. A primal-dual solution strategy is implemented for support vector based classification as a proof of concept. A modification of the growing neural gas algorithm as an efficient tool for gene expression profiling is reported. The modified algorithm has been successfully used for class discovery and class prediction of three previously reported gene expression data sets from the literature. A novel algorithm, namely, adaptive centroid algorithm (ACA), is developed for single as well as multi cluster gene assignments. It uses an analysis of variance based performance criterion, analogy to center-of-mass principle for heterogeneously distributed mass elements, Euclidian distances, and the given gene expression data set to give unique solutions. ACA has been employed for the analysis of two colon cancer case studies. Time- and strain-specific genetic alterations during different stages of colon tumorigenesis following azoxymethane (the DNA alkylating agent that induces tumor formation in mice) treatment are reported in the first case study. Lesion-specific molecular alterations, which may provide insights into the underlying mechanisms of chronic ulcerative colitis-related tumorigenesis in humans, are observed in the second case study. ^