The human body at its genetic core consists of enzymes
that
catalyze
reactions. These reactions generate compounds by consuming a set of
other compounds. Chemical molecules (drugs) that inhibit enzymes can
reduce or increase the generation of one or more compounds. The goal of
drug discovery is to derive chemical molecules that inhibit enzymes to
have a therapeutic effect by increasing or reducing the generation of
more or more target compounds while minimizing the negative effects
(side effects) on other important compounds. Currently this is done
mainly by testing a large number of molecules to inhibit enzymes using
high throughput screening. This is a very costly (more than US $800
million per drug) and inaccurate (90 to 95% of the drugs entering the
clinical development fail) process. New strategies to reduce the
attrition rates of drugs by considering toxicity as well as efficacy
prior to clinical study is sorely needed. Metabolic networks define the
relationships between enzymes and compounds. Although many models for
metabolic networks have been developed, they have not been used for
predicting the efficacy and the side effects of drugs.
We are developing effective computational models and
tools
that
will exploit the structure of the metabolic networks to better
understand the impact of molecules on different enzymes as well
understanding which enzymes should be inhibited so that the target
compounds can be appropriate affected. The proposed work can be broadly
divided into the following two tasks:
Development of biologically accurate computational
cost
models for metabolic networks.
Development of efficient algorithms for identifying
target
enzymes that can have a desired effect on target compounds. Given a set
of target (unwanted) compounds, we will develop computational tools
that will identify a subset of enzymes whose inhibition disrupts the
production of the target compounds with minimum or close to minimum
side effect (i.e., disruption of non-target compounds) as defined by
the underlying cost model.
We will validate the accuracy of the proposed models and
the
search
methods using successful and failed drugs that have been screened in
existing or ongoing preclinical trials.
Publications:
Padmavati Sridhar, Tamer Kahveci, Sanjay Ranka,An
iterative
algorithm for metabolic network-based drug target identification,
Proceedings of PSB 2007. (Abstract)
(PDF)