Cell Signaling Networks From the Bottom Up Anthony M.L. Liekens BioModeling and BioInformatics Anthony M.L. Liekens BioModeling and BioInformatics

ESIGNET European NEST project with Birmingham, Dublin, Jena Signal transduction pathways Black box models of conceptual networks Computational properties? Evolvability? European NEST project with Birmingham, Dublin, Jena Signal transduction pathways Black box models of conceptual networks Computational properties? Evolvability?

Signal Transduction Most proteins known for metabolic processes, cell maintenance Many proteins responsible for transduction of signals information processing Estimated 5% of human genes Elementary and common motif: Phosphorylation cycle

Phosphorylation Cycle Phosphorylating kinase Dephosphorylating phosphatase

Chemical “Transistor” Kinase concentration = input Equilibrium concentration of E-P: Phosphorylation acts as switch

Signaling Networks Phosphorylation cycle is elementary motif that acts as transistor Phosphorylated protein catalyzes other phosphorylations Cascading networks of cycles allow for the implementation of “computations” Small example: Chemotaxis

Chemotaxis of E. coli (1) Receptors sample environment Chemotaxis controls actuators Cell moves to higher concentrations in nutritional gradient (Bray et al, Computational Cell Group, University of Cambridge)

Chemotaxis E. coli (2) Signaling network for chemotaxis in E. coli

Higher Organisms Networks may compromise >80 kinases and phosphatases (Gomperts et al, Signal Transduction, 2002) Increasing complexity and feedback ➡ hard to infer knowledge Numerous applications (Kitano, Science, 2000) Responses to inflammation

Modular Approach Recognize small, common motifs ➡ behavior is mathematically comprehensive Replace motif by “super node” that acts similarly Hierarchical integration leads to understanding of complex networks (Kholodenko et al, FEBS Letters, 1995; Weng et al, Science, 1999; Hartwell et al, Nature, 1999; Kholodenko et al, Topics in Current Genetics, 2005)

Observed Behaviors Boolean operations and simple binary computations Integration and amplification of signals Bandpass frequency and noise filters Bistable switches, oscillators and hysteresis through feedback Neural networks (Wolf and Arkin, Current Opinion in Microbiology, 2003) Related body of work in gene expression

Bottom-up Approach Construct conceptual motifs from the bottom up, rather than dissecting real networks from the top down What elementary mathematical operations can be represented as reaction networks? What kind of functions can we construct out of these? Are these networks “evolvable”?

Elementary Motif A catalyzes production of X, (rate constant k1) with abundant resources X decays (k2) to waste ODE model with mass-action kinetics If k1 = k2, [ X ] = [ A ] in equilibrium

Elementary Algebraic Operations Addition Multiplication Subtraction Division nth Root

Complex Computations Elementary operations can be combined Output of one network serves as the input of the next network Second network does not influence first, but is dependent on it Equilibrium state = composed function Allows more complex computations

Example: ABC Formula “Solves”

Example: Polynomial Network computes

Algebra of phosphorylation cycles? ?

Ongoing Research Behavior of elementary operations, dropping assumptions Feedback mechanisms In silico evolution of such networks Stochastic models Molecular dynamics simulations Verification of signaling networks Bring understanding to real problems

People Involved Peter Hilbers (PI) Huub ten Eikelder (UD) Dragan Bosnacki (UD) Anthony Liekens (Postdoc) Marvin Steijaert (AiO) Harm Buisman (thesis, finished) Jeroen van den Brink (thesis) Sander Allon (internship) Sjoerd Crijns (internship)

Questions?