Home / People / Directory / Jeremy Gunawardena

Jeremy Gunawardena


BSc Mathematics, Imperial College, 1977
PhD Algebraic Topology, Cambridge University, 1981

Gunawardena lab website

Research summary

We are developing mathematical and computational techniques for analysing cellular physiology from a molecular perspective. This has taken two paths, one mechanistic, starting from knowledege of molecular networks, the other Bayesian, starting from -omic data and trying to infer molecular interactions. We have a broad interest in understanding how to integrate these approaches.

On the mechanistic path, we have been studying enzyme bifunctionality. Enzymes are renowned for being highly specific but, in some cases, the same protein may catalyse two reactions. Such bifunctionality appears to fall into two classes, one in which the two reactions are post-translational modification and demodification and the other, in which the reactions are adjacent, or nearly adjacent, steps in a metabolic pathway. Example of the first class are the EnvZ/OmpR two-component osmoregulatory system in E coli and the phosphofructokinase2-fructose-2,6-bisphosphatase (PFK2-F2,6BPase) regulator in mammalian glycolysis. In previous work from our laboratory, we have developed new methods of steady-state analysis (PMID: 17704153, 18849417) that can be applied to the biochemical networks underlying the interaction of these enzymes with their multiple substrates and products and have applied these to PFK2-F2,6BPase, [1]. Our work shows that the bifunctionality gives rise to a more general form of robustness in the state of the modified substrate to that found previously for EnvZ/OmpR (PMID: 18077424). This suggests, at least for this class of bifunctional enzymes, that bifunctionality has a general role in providing robustness to changes in enzyme or substrate concentrations. We are developing more powerful mathematical methods to try and exhibit this for all modification-bifunctional enzymes, [2].

We are also investigating the other class of bifunctional enzymes, using as a guiding example dihydrofolatereductase (DHFR) and thymidylatesynthase (TS). These enzymes catalyse key steps in amino acid and nucleotide biosynthesis. They are joined in a bifunctional enzyme (DHFR-TS) in parasites like Plasmodium falciparum in malaria but remain as separate monofunctional enzymes in humans. This makes DHFR-TS an important anti-malarial drug target, for which an understanding of the selective advantage arising from bifunctionality could be of significant clinical importance. Unlike modifiction-biufunctionality, which regulates the modification state of metabolic substrates, this class of bifunctional enzymes seems to regulate flux. We believe that, as before, the bifunctionality confers some form of robustness. However, the biochemical details are substantially more complex than for PFK2-F2,6BPase, in part because of additional co-factors and random-order substrate binding and product release. We are using computational methods to compare the bifunctional and separated enzymes and also developing more powerful mathematical tools for analysing the corresponding enzymatic networks. Our overall aim is to work out the role of enzyme bifunctionality in the control of metabolism.

On the Bayesian path, we have developed a collaboration with the reproductive endcrinology lab of Dr Nicholas Orsi at the University of Leeds. Experimental assays are now available for rapidly measuring a broad panel of hormones and cytokines. We want to elicit from such data a molecular network view of how mammalian reproduction is regulated. We have used Bayesian network methods to analyse circulating prolactin, steroid hormone and cytokine levels during murine lactation, [3]. This is a phase of the mammalian reproductive cycle in which the role of cytokines has not yet been clarified.  We have developed manual curation and automated methods to generate seed networks and used soft constraints to improve computational efficiency for finding near-optimal Bayesian networks. Our results suggest a reassessment of prolactin as a key regulator during this phase. Among the potential clinical impacts of this work are improved diagnostic methods during human pregnancies but the longer-term goal is to stimulate a shift in thinking within the field of endocrinology from a component-based to a network-based perspective. Ultimately, we hope to understand and exploit the synergy between mechanistic and Bayesian methods.

©2019 Cell Decision Process Center all rights reserved

This page last modified on May 19th, 2011