1. University of Illinois atUrbana-Champaign BIOINFORMATICS ON NETWORKS Nick Sahinidis Chemical and Biomolecular Engineering

2. MOTIVATION • Genomics and proteomics help us understand the structure, properties, and function of single genes and proteins • Genes and proteins function in complex networks • Bioinformatics on biochemical networks aims to understand and rationally manipulate networks of genes and proteins • These networks are very complex • http://www.expasy.org/cgi-bin/show_thumbnails.pl • http://www.expasy.org/cgi-bin/show_thumbnails.pl?2 • http://www.genome.ad.jp/kegg/pathway.html

3. LEARNING OBJECTIVES (two lectures) • Introduction to: • Metabolic networks • Flux balance analysis • S-systems theory • Gene additions and deletions • Pathway reconstruction from data

4. METABOLIC NETWORKS • Definitions • Metabolic network: a system of interacting proteins and small molecules converting raw materials to energy and other useful substances in a living organism • Metabolites: materials consumed or produced in a metabolic network • Enzymes: proteins that catalyze reactions • The sets of metabolites and enzymes of a network are not necessarily disjoint • Key observation • A large proportion of the chemical processes that underlie life are shared across a very wide range of organisms

5. GRAPHICAL REPRESENTATION • Nodes represent metabolites and enzymes • Arcs correspond to reactions and modulation • Dotted or colored lines often reserved to denote modulation • A negative sign associated with an arc is used to denote inhibition

6. METABOLIC NETWORK EXAMPLE A B C E D • Five metabolites (A, B, C, D, E) • Six reactions (one reversible and five irreversible) • Network interacts with environment through: • Consumption of A • Secretion of E • Consumption or secretion of C and D

7. FLUX BALANCE ANALYSIS • Pseudo steady-state hypothesis: metabolic dynamics are much faster compared to those of the environment • Model network through steady-state mass balances for metabolites • For each metabolite, its rate of consumption must equal its rate of production

8. Internal Fluxes v1: A B v2: B C b2 v3: B D v4: D B v2 v1 v6 v5: C D b1 b4 v4 v5 v6: C E v3 v7 v7: 2D E Exchange Fluxes Network Boundary b1: A b3 b2: C b3: D b4: E FBA EXAMPLE A B C E D Exchange fluxes may be positive (system output) or Negative (input to metabolic network)

9. b2 v2 v1 v6 b1 b4 v4 v5 v3 Steady state mass balances v7 A: - v1 - b1 = 0 B: v1 + v4 – v2 – v3 = 0 Network Boundary b3 C: v2 - v5 - v6 - b2 = 0 D: v3 + v5 - v4 - 2v7 - b3 = 0 E: v6 + v7 - b4 = 0 FBA EQUATIONS A B C E D Sign restrictions 0  v1,…,v7 b1  0 -  b2  + -  b3  + b4  0

10. MODELING WITH FBA • Problem #1: Interpret metabolic network behavior • Hypothesis: Network is an optimizer • Likely objectives: • Maximize growth • Minimize energy consumption • Leads to a linear program • Problem #2: Manipulate a metabolic network to produce certain desired products through • Control of external fluxes • Structural manipulations in the network

11. GENE ADDITIONS AND DELETIONS • Two-level problem • Upper level: maximize a bioengineering objective through gene knockouts • Lower level: cell is still an optimizer that seeks to optimize its own objective through adjusting internal fluxes • Use binary variable for each gene to decide whether to knock it out or not (or whether to over-express) • Inner linear program can be converted to a set of linear equalities and inequalities via duality theory giving rise to a mixed-integer linear program for the overall problem

12. REFERENCES AND FURTHER READING • B. Palsson, 2000 Hougen Lectures • http://gcrg.ucsd.edu/presentations/hougen/hougen.htm • E. Voit, Computational Analysis of Biochemical Systems, Cambridge University Press, 2000. • N. Friedman, Inferring cellular networks using probabilistic graphical models, Science, 303, 799-805, 2004. • Metabolic Systems Engineering course: • http://archimedes.scs.uiuc.edu/courses/meteng.html