Prof. Gregory N Stephanopoulos

Research Interests: Metabolic and Biochemical Engineering, Cell Culture, Biotechnology

Applications of genetic engineering to strain improvement for metabolite overproduction have been slow in coming, due partly to the complexity of metabolic networks, and also to the lack of a general approach to metabolic engineering. In order to address this problem, our group has focused on developing methods for the measurement and control of metabolic fluxes in-vivo. These methods allow the systematic study of cellular responses to genetic and environmental perturbations, and the rational design of metabolic modifications. This collaborative research effort aims at the development of a framework of metabolic engineering within which directed alterations of intracellular flux distributions are attempted as a means of optimizing cellular function and product synthesis. Along with fermentor control, continuous cultures, genetic stability of auxotrophic and recombinant strains, and cell immobilization are also studied.

Another interest in our research is technologies for the large-scale cultivation of mammalian cells. Mammalian cells are the preferable system for the production of natural, but until recently unavailable, potent biomolecules in a form that ensures high activity and stability. Current research focuses on the development of novel high-cell density and productivity bioreactors, but also on the elucidation of the effect that the bioreactor environment has on the amount and quality of product formed. In this context, intracellular processing steps (such as folding, glycosylation, and secretion) participating in the post-translational modification of a protein are studied. We also study cell death in order to maximize reactor productivity by maximizing the number of viable cells in culture.

During the past 10 years, we have pursued methodologies for the upgrade of data collected during a fermentation. In addition to algorithms embodying research in state estimation, control, data consistency and reconciliation, this research has led to the realization that there is a good deal of information in fermentation data that cannot be extracted by the usual mechanistic models. Current research is investigating techniques from pattern recognition and artificial intelligence, to be incorporated in a real time workstation which will utilize data for post-mortem process analysis and rule extraction, as well as fermentation diagnosis and supervisory control.

Linking the expression phenotype to the metabolic phenotype

Metabolic and biochemical engineering -- Applications of genetic engineering to strain improvement for metabolite overproduction have been slow in coming, due partly to the complexity of metabolic networks, and also to the lack of a general approach to metabolic engineering. In order to address this problem, our group has focused on developing methods for the measurement and control of metabolic fluxes in-vivo. In parallel, we are measuring transcriptional profiles using gene arrays. These methods allow the systematic study of cellular responses to genetic and environmental perturbations, and the rational design of metabolic modifications. This collaborative research effort aims at the development of a framework of metabolic engineering within which directed alterations of intracellular flux distributions are attempted as a means of optimizing cellular function and product synthesis.

Bioinformatics and pattern discovery -- During the past 12 years, we have pursued methodologies for upgrading the information content of biological and bioprocess data - "data mining." Conventional mechanistic approaches to modeling, while successful in many physical and chemical systems, have limited applicability in complex biological systems. This has been primarily due to the lack of analytical sensors for non-invasive probing of the cellular state and function, as well as the considerable complexity of biological systems. The approaches that we have developed are primarily data driven and seek to take advantage of the structure of available data by analyzing and searching historical databases for distinctive patterns. The developed pattern discovery and multivariate statistical approaches have been consolidated in a flexible, user friendly, software package called dbminer. It has been applied to the analysis of data from bioprocesses, solid phase peptide synthesis, NIR measurements of vaccine production, casein content of milk samples, and differential gene expression.