How Can We Understand Metabolism?

David Fell

When genome sequencing projects started to gather pace, many biochemists and molecular biologists regarded metabolism as a largely solved problem, at least for the biochemists' usual systems of study (Escherischia coli, yeast and the rat). In biochemistry textbooks, before they entered on specific details of metabolism, authors elaborated general principles, such as: control by rate-limiting steps; control of pathway rate by feedback inhibition; control of pathways by feedback inhibition on the first committed step after a branch point, etc. Yet we now know that this confidence in these principles was misplaced. Attempts to manipulate metabolism in specific ways by changing the expression levels of enzymes have usually not worked as anticipated. Furthermore, many of the supposed principles have been shown by theoretical analysis to be wrong, or not general. Although the theoretical developments that have led to a re-evaluation of the principles (Metabolic Control Analysis and Biochemical Systems Theory) provide frameworks for a more rigorous analysis and understanding of metabolism, they do not lead to universally applicable laws of metabolism.

Thus although we are certain that the links connecting the genome to the metabolic phenotype are clearly that genes specify enzymes and the enzymes together catalyse a set of reactions that taken together constitute the metabolic network, this does not allow us to make accurate predictions about the metabolic responses even of those organisms for which we know the characteristics of many of the genes and enzymes. One solution that is being proposed to this problem is to construct computer simulations of the whole metabolism of a cell. If we look forward to the time when the practical problems have been overcome, and such simulations are running and making verifiable predictions, will we be able to say that we understand metabolism better at that point? It is clear that any such simulation will involve thousands of variables, parameters and non-linear functions, so that the whole model will be too large for any biochemist to hold in mind. If such simulations are to be more than in silico replicas of the real cells, how will we use them to promote understanding of metabolism?