…

(En) Part I: Basic

References

1. Ingalls, B. P. (2013). Mathematical modeling in systems biology: an introduction. MIT press.
2. Strogatz, S. H. (2018). Nonlinear dynamics and chaos with student solutions manual: With applications to physics, biology, chemistry, and engineering. CRC press.

Stability of Fixed Points

For this part, I mainly talk on nonlinear gene networks. Check (En) Systems Biology for the context and examples.

Take a nonlinear network of species interaction with each other, the dynamics of the system consists of species are described as a system of ODEs. How to mathematically solve (or describe) them? For this part, check Strogatz’s book (ref 2).

One-Dimensional Case

$\frac{dx(t)}{dt}=f(x(t))$ is an ODE. How to solve it is described in 微积分. As we can see, we can only solve some very specific ODEs analytically. If we can’t solve analytically, it is helpful to sketch the trend of the evolution without a lot of detail.

If the solution of $\frac{dx(t)}{dt}=0$ is $x=x_i$ (if nonlinear, there can be multiple solutions, and we call them fixed points), then \(x=x_i\) might be a candidate that finally the system will rest at this point. Will this happen?

We focus on the local part near the fixed points. Place your state near a fixed point, will it go to the fixed point or diverge further away? And then, can we sketch the how phase plane by this?

(Ingalls’ 4.2) The long-time (i.e., asymptotic) behavior of biochemical and genetic networks will be either

• convergence to a steady state; or
• convergence to a sustained periodic oscillation, referred to as limit-cycle oscillation. (Shi’s comment: sorry for the messy structure but this only works for multi-dimensional case)

Other dynamic behaviors (divergence and chaos, for example) do not often occur in systems biology models.

Keep only the first-order term, we have $\frac{dx(t)}{dt}\vert_{x=x_i}=f’(x_i)(x(t)-x_i)$. As the (Example 2.2.1) shown in Strogatz’s book ($f(x)=x^2-1$), the stability of fixed points are determined by the symbols of $f(x_i)$. If negative, go to the fixed point (stable); if positive, diverge away (unstable). More tricky case that $f(x_i)=0$, see Strogatz’s book, not this post.

Multi-dimensional Case

For the example, check Ingalls’ model (4.2). For a general expression, $\mathbf{x(t)}=(x_1(t),…,x_n(t))^{T}$ is an \(n\times 1\) vector, and $\mathbf{f}(\mathbf x(t))=(f_1(x_1(t)),…,f_n(x_n(t))^T$ is an $n\times 1$ vector, the system of ODEs is therefore $\frac{d}{dt}\mathbf{x}(t)=\mathbf{f}(\mathbf x(t))$. Fixed points are $\mathbf x_i$. Consider the local properties near the fixed points, only keep the first-order terms, we have $\frac{d}{dt}\mathbf{x}(t)\vert_{\mathbf x=\mathbf x_i}=\mathbf{J}(\mathbf x_i)\cdot (\mathbf x(t)-\mathbf x_i)$, where $J_{mn}(\mathbf x_i)=[\frac{\partial f_m(\mathbf x)}{\partial x_n}]\vert_{\mathbf x=\mathbf x_i}$ is the Jacobian matrix.

As in 1D case, the stability is determined by the symbols of the first derivatives of $f(x)$ at the fixed points, what criteria we use in the multi-dimensional case? The eigenvalues of the Jacobian matrix at the fixed points!

Read Ingalls’ 4.2 for the great mathematical explanation. To summarize,

Linearized Stability Criterion

• If both eigenvalues of the Jacobian have negative real part, then the state is stable.
• If either eigenvalue has positive real part, then the steady state is unstable.

… not addressed the case of eigenvalues with zero real part … rarely encountered…

Bifurcation

Check Strogatz’s related sections.

If we change the value of a parameter (parameters) in ODEs, the structure of the flow diagram can change. The number and the stability of fixed points can change. Qualitative changes in the dynamics are called bifurcations.

Example: Saddle-node bifurcation, transcritical bifurcation and pitchfork bifurcation, hopf bifurcation…

(Zh) Part II: Intermediate

References

1. 梁昆淼，高等教育出版社《理论力学（第四版）》，第十章。

线性 $f$：$f(ax_1+bx_2)=af(x_1)+bf(x_2)$。

物理学中的叠加原理是系统具有线性性质的体现。在该情况下，系统的整体性质可以由组成该系统的所有子系统的性质叠加尔得到，这也表明各子系统之间是没有相互作用的，或者这种相互作用对所讨论的性质而言是可以忽略不计的。

微扰法

用小参数展开进行主机近似的方法。

逐级近似—— e.g., 非线性振动的求解。求越来越高阶的振动方程，每次求解的时候舍弃更高阶小量，然后逐级求解。

参数共振—— e.g., 秋千越荡越高。共振不是通过驱动力实现的，而是通过参数的变化而实现的。参数共振与受迫振动的不同。