The Paradox of Chemical Reaction Networks : Robustness in the face of total uncertainty By David Angeli: Imperial College, London University of Florence, Italy Definition of CRN List of Chemical Reactions: 11S1 12 S 2 1n S n 11S1 12 S 2 1n S n 21S1 22 S 2 2 n S n 21S1 22 S 2 2 n S n r1S1 r 2 S 2 rn S n r1S1 r 2 S 2 rn S n The Si for i = 1,2,...,n are the chemical species. The non-negative integers , are the stoichiometry coefficients. Example of CRN E + S0 ES0 E + S1 ES1 E + S2 F + S2 FS2 F + S1 FS1 F + S0 F FS1 S0 FS2 S1 ES0 S2 ES1 E Discrete Modeling Framework Stochastic: Discrete event systems: PETRI NETS Reaction rates: mass-action kinetics F FS1 S0 FS2 S1 ES0 S2 ES1 E Problem : Markov Chain with huge number of states Continuous Modeling Framework Deterministic: Continuous concentrations, ODE models Large molecule numbers: variance is neglegible Isolated vs. Open systems • Thermodynamically isolated systems: Reaction rates derived from a potential. Every reaction is reversible. Steady-states are thermodynamic equilibria: detailed balance Passive circuits analog of CRNs. Entropy acts as a Lyapunov function. • Open systems: Some species are ignored: clamped concentrations. Partial stoichiometry. Arbitrary kinetic coefficients. No obvious Lyapunov function. Possibility of “complex” behaviour. Relating Dynamics and Topology • How does structure affect dynamics ? • How robust is the net to parameter variations ? • Does the reaction converge or oscillate ? Qualitative tools: can work regardless of specific parameters values. • How to define robustness ? Consistent qualitative behavior regardless of Parameters or kinetics. MAPK random simulation More random simulations What is Persistence • Notion introduced in mathematical ecology: non extinction of species • For positive systems x f (x) it amounts to: lim inf xi (t ) 0 i t •For systems with bounded solutions equivalently: ( x) R n 0 Petri Nets Background F FS1 S0 FS2 S1 ES0 S2 Bipartite graph: PLACES (round nodes) TRANSITIONS (boxes) ES1 E P-semiflow: non-negative integer row vector v such that vS=0 T-semiflow: non-negative integer column vector v with Sv=0 Support of v: set of places i (transitions) such that v_i>0 Incidence matrix = Stoichiometry matrix = S Necessary conditions for persistence •Let r(x) denote the vector of reaction rate •We assume that for x>>0, r(x)>>0 •Under persistence, the average of r(x(t)) is strictly positive and belongs to the kernel of S •Hence, Persistence implies existence of a T-semiflow whose support coincides with the set of all transitions. This kind of net is called: CONSISTENT Petri Net approach to persistence SIPHON: Input transitions Included in Output transitions F FS1 S0 FS2 S1 ES0 S2 ES1 E Assume that x(tn) approaches The boundary. Let S be the set of i such that xi(tn) 0 Then S is a SIPHON Structurally non-emptiable siphons A siphon is structurally non-emptiable if it contains the support of a positive conservation law F FS1 S0 P-semiflows: E+ES0+ES1 F+FS2+FS1 S0+S1+S2+ES0+ES1+FS2+FS1 FS2 S1 S2 Minimal Siphons: ES0 ES1 E { E, ES0, ES1 } { F, FS2, FS1 } { S0, S1, S2, ES0, ES1, FS2, FS1 } All siphons are SNE PERSISTENCE Network compositions Full MAPK cascade 22 chemical species 7 minimal siphons 7 P-semiflows whose supports coincide with the minimal siphons Hopf’s bifurcations • Symbolic linearization: x f (x) f x x x • Characteristic polynomial f det( sI ) ( s 6 a1s 5 ... a6 ) s 3 x • Hurwitz determinant Hn-1 = 0 is a necessary condition for Hopf’s bifurcation (n=6). Hurwitz determinant a1 1 H5 0 0 0 a3 a2 a1 1 0 a5 a4 a3 a2 a1 0 a6 a5 a4 a3 0 0 0 a6 a5 • ai are polynomials of degree i in the kinetic parameters • det(H5) is a polynomial of degree 15 in the kinetic parameters (12 parameters + 5 concentrations) • Number of monomials is unknown • Letting all kinetic constants = 1 except for k1 k3 k5 k7 yields 68.425 monomials all with a + coefficient Remarks • This is much stronger than: det(Hn-1) is positive definite. • Purely algebraic and graphic criterion for ruling out Hopf’s bifurcations expected. • Notion of negative loop in the presence of conservation laws. Conclusions • CRN theory: open problems and challenges • At the cross-road of many fields: - dynamical systems - biochemistry - graph theory - linear algebra HAPPY 60 EDUARDO