Limit Cycles II. Prof. Ned Wingreen MOL 410/510. How to prove a closed orbit exists?

Size: px

Start display at page:

Download "Limit Cycles II. Prof. Ned Wingreen MOL 410/510. How to prove a closed orbit exists?"

Leslie Davis
5 years ago
Views:

1 Limit Cycles II Prof. Ned Wingreen MOL 410/510 How to prove a closed orbit eists? numerically Poincaré-Bendison Theorem: If 1. R is a closed, bounded subset of the plane. = f( ) is a continuously differentiable vector field on an open set containing R 3. R does not contain any fied points 4. There eists a trajectory C that is confined in R Then either C is a closed orbit or it spirals towards a closed orbit as t. Either way, R contains a closed orbit. P C R Figure 1: Poincaré-Bendison Theorem Eample Glycolytic oscillations Background Organisms may obtain energy by breaking down sugar. Glycolysis can proceed in an oscillatory fashion. In a simple model with = concentration of ADP (adenosine diphosphate) y = concentration of F6P (fructose-6-phosphate) 1

2 we have ẋ = + ay + y ẏ = b ay y y F6P ADP Figure : Glycolysis model Starting from Kinetic Equations A brief aside on deriving dimensionless equations from kinetic equations - by rescaling chemical concentrations and time by appropriate units, one can generally reduce the number of parameters. In our case, the underlying kinetic equations are d[a] = µ[a] + α[f] + γ[a] [F] dt d[f] = β α[f] γ[a] [F] dt which has four (dimensionful) parameters, α, β, γ, and µ. We ll start by rescaling time. Divide by µ, which is a rate, to give d[a] d(µt) = d[a] dt = [A] + (α/µ)[f] + (γ/µ)[a] [F] d[f] d(µt) = d[f] dt = β/µ (α/µ)[f] (γ/µ)[a] [F], where we have defined t = µt. Now, to eliminate the parameter γ/µ in front of the last term in each equation, while continuing to measure [A] and [F] in the same units, we ll rescale the concentrations [A] and [F] as which yields [A] = (µ/γ) 1/ [F] = (µ/γ) 1/ y, 1/ d (µ/γ) dt = (µ/γ)1/ + (α/µ)(µ/γ) 1/ y + (γ/µ)(µ/γ) 3/ y (µ/γ) 1/ dy dt = β/µ (α/µ)(µ/γ)1/ y (γ/µ)(µ/γ) 3/ y,

3 and simplifies to ẋ = + (α/µ)y + y ẏ = (γ/µ) 1/ (β/µ) (α/µ)y y. These are our original, dimensionless equations, and now we can see eactly how the two remaining dimensionless constants a and b depend on the underlying rates in the kinetic equations: a = α/µ and b = (γ/µ) 1/ (β/µ). Intuition Since rate of F6P ADP increases with ADP, we can get overshooting, i.e. F6P gets depleted, ADP has no source so it also gets depleted, followed by slow recovery of F6P. This is a possible oscillator, but how can we prove a limit cycle? Find the nullclines y ẏ = 0, y = b a+ ẋ > 0, ẏ < 0 ẋ > 0, ẏ > 0 P ẋ = 0, y = a+ ẋ < 0, ẏ > 0 ẋ < 0,ẏ < 0 Figure 3: Nullclines sketch ẋ = 0 = + ay + y = 0 = y = a + ẏ = 0 = b ay y = 0 = y = b a + 3

4 b a+b. Solve for fied point: = b, y = Does this prove a limit cycle? NO! We could have a stable fied point at P, or trajectories could spiral out to. Indeed, at the intersection of the nullclines ẋ = ẏ = 0, so P is a fied point. Fied point We can t apply Poincaré-Bendison yet because of the fied point. We can use P-B, though, if the fied point is a repeller, because then our trapping region is just R \ P. Analyze stability of fied point P by linearizing the differential equations around the fied point: ( ) ẋ ẋ ( ) 1 + y a + Jacobian A = ẏ y ẏ y = y (a + ) Fied point P : = b,y = b a + b ) ( 1 + b A(P) = a+b a + b b a+b (a + b ) = det A(P) = a + b > 0 τ = tr A(P) = 1 + b a + b (a + b ) = b4 + (a 1)b + (a + a ) a + b In general, we can quickly determine the stability of a fied point if we know and τ, i.e. the determinant and trace of the Jacobian at the fied point, because the eigenvalues are given by λ J = 1 (τ ± τ 4 ). (It s worth remembering that for any square matri = Π i λ i and τ = Σ i λ i.) saddle τ unstable nodes unstable spirals centers stable spirals stable nodes Figure 4: Stability and types of fied points The fied point P is unstable for τ > 0, stable for τ < 0. The dividing line τ = 0 is at b = 1 (1 a ± 1 8a) 4

5 b 1 stable limit cycle stable fied point a Figure 5: Glycolysis analysis y b/a (b, b/a) R? slope = 1 Find a trapping region Figure 6: Trapping region Eamine Figure 6 (page 6). The vectors for ẋ = 0 or ẏ = 0 follow from the last figure. But what about the circled vector? 5

6 Circled vector is trapping if ẏ < ẋ (i.e. ẋ + ẏ < 0) along the boundary: ẋ + ẏ = + b = ẋ + ẏ < 0 if > b, so the dashed lines do define a trapping region. A model for glycolysis We can conclude that our glycolytic model functions as in Figure 5 (page 5).Does this make sense? If a is too big, then F6P ADP even for low levels of ADP, so there s no chance for a pool of F6P to accumulate. At a fied a, if b is too small then new F6P will instantly turn into ADP and get used up, so the system is locked in a low flu state. If b is too big, ADP will never be low enough, so the system is locked into a high flu state. 6

7 How to characterize a limit cycle? (nearly) harmonic oscillator vs relaation oscillator period amplitude Eample - van der Pol oscillator ẍ + µ( 1)ẋ + = 0 Damped harmonic oscillator: ordinary damping for > 1, negative damping for < 1. Large amplitude oscillations will decay, but small amplitude oscillations will get pumped up. Like a parent pushing a child on a swing... It can be proven that the van der Pol oscillator has a single, stable limit cycle for each µ > 0. van der Pol as a relaation oscillator (µ >> 1) ẍ + µ( 1)ẋ = d dt [ ẋ + µ( 1 ] 3 3 ) The van der Pol equation implies Let F() = 1 3 3,ω = ẋ + µf(). ω =, ẋ = ω µf(), and if we let y = ω/µ, ẋ = µ[y F()] ẏ = 1 µ fast slow Nullclines So there are two separated timescales: Period crawls t µ jumps t 1/µ The period of relaation oscillator is dominated by crawls. For van der Pol, by symmetry T t B t A dt. On the slow branches: dy dt dy d d Nullcline dt = df() d d dt = ( 1) d dt. But dy dt = 1 µ, 7

8 y ẋ = 0, y = F() = D fast: ẋ µ A /3 ẏ = 0 slow: ẏ 1 µ 1 1 C /3 B Figure 7: van der Pol nullclines so and Therefore: 1 µ ( 1) d dt dt µ( 1) d. tb B T = dt µ( 1) d t A A [ ] = µ ln 1 = µ(3 ln ) µ since ( A =, B = 1) 8

Systems Biology: A Personal View XX. Biological oscillators: Hopf Bifurcation & Glycolysis. Sitabhra Sinha IMSc Chennai

Systems Biology: A Personal View XX. Biological oscillators: Hopf Bifurcation & Glycolysis Sitabhra Sinha IMSc Chennai www.bio.brandeis.edu Oscillations in the Chlorine dioxide-iodine- Malonic Acid (ClO