A | B | A | B | ||
$x_1$ (CDH1)} | 1.7924 | 0.1111 | $x_4$ (ZEB2)} | 0.0170 | 1.8169 |
$x_2$ (SNAI1)} | 0.3359 | 3.2229 | $x_5$ (miR-200)} | 1.7924 | 0.1111 |
$x_3$ (ZEB1)} | 0.0522 | 1.8224 | $x_6$ (miR-203)} | 2.9873 | 0.1851 |
The control of complex nonlinear dynamical networks is an ongoing challenge in diverse contexts ranging from biology to social sciences. To explore a practical framework for controlling nonlinear dynamical networks based on meaningful physical and experimental considerations, we propose a new concept of the domain control for nonlinear dynamical networks, i.e., the control of a nonlinear network in transition from the domain of attraction of an undesired state (attractor) to the domain of attraction of a desired state. We theoretically prove the existence of a domain control. In particular, we offer an approach for identifying the driver nodes that need to be controlled and design a general form of control functions for realizing domain controllability. In addition, we demonstrate the effectiveness of our theory and approaches in three realistic disease-related networks: the epithelial-mesenchymal transition (EMT) core network, the T helper (Th) differentiation cellular network and the cancer network. Moreover, we reveal certain genes that are critical to phenotype transitions of these systems. Therefore, the approach described here not only offers a practical control scheme for nonlinear dynamical networks but also helps the development of new strategies for the prevention and treatment of complex diseases.
Citation: |
Figure 1. A schematic illustration of the domain control. (A) The nonlinear networked system can exhibit transitions between the different stable steady states. The red solid, blue solid and green dashed lines represent one stable steady state (attractor), another stable steady (another attractor) state and unstable states, respectively. The green dashed line with an arrow indicates the transition from one attractor to another attractor. (B) and (C) provide illustrations of the concept of domain control, namely the transition from the domain of attraction of one attractor to the domain of attraction of another attractor. (B) The domain of attraction of one attractor (corresponding to the red solid line in (A)). Each state in the domain of attraction is represented by a red ball. The horizontal axis is a projected state-space and the vertical axis is the potential, which indicates the relative instability of individual states. (C) The domain of attraction of another attractor (corresponding to the blue solid line in (A)). Each state in the domain of attraction is represented by a blue ball
Figure 2. The EMT core network and domain control strategies for its phenotype transition. (A) Diagram (redrawn from ^{[41]}) of the EMT core network. The arrows and short bars represent activation and inhibition, respectively. For example, SNAI1 activates ZEB1 but inhibits miR-203 expression. (B) The practically required driver nodes (genes) for realizing domain controllability of this network. Mesenchymal cells can be induced from epithelial cells and back by any one of the four genes (SNAI1, ZEB1, ZEB2 and miR-203) and any one of the two genes (SNAI1 and miR-203), respectively
Figure 3. Domain control of the EMT core network transition from the epithelial state to the mesenchymal state. The representative node state trajectories $x(t)$ (A-B), control function $u(t)$ (C) and control error $e(t)$ (D) are depicted when the practically required driver node, namely miR-203, is controlled. The symbols $x_2$ and $x_6$ indicate the activities of SNAI1 and miR-203, respectively. The black dashed and blue solid lines represent the initial epithelial state (i.e., the undesired state with no control) and desired mesenchymal state, respectively
Figure 4. The Th differentiation cellular network and domain control strategies for its phenotype transition. (A) Diagram (redrawn from the reference ^{[39]}) of the Th differentiation cellular network. (B) The practically required driver nodes (genes) for realizing domain controllability of this network. Th1 and Th2 cells can be induced from Th0 cells by any one of the five genes (T-bet, IFN-$\gamma$, IFN-$\gamma$R, JAK1 and STAT1) and any one of the four genes (GATA-3, IL-4, IL-4R and STAT6), respectively
Figure 5. Domain control of the Th differentiation cellular network transitions from Th0 to Th1 and Th2. The representative node state trajectories $x(t)$ (A, B), control function $u(t)$ (C) and control error $e(t)$ (D) are depicted. The symbols $x_1$ and $x_{22}$ indicate the activities of GATA-3 and T-bet, respectively. The black dashed line indicates the Th0 phenotype, i.e., the initial state with no control. The red line with circle markers and blue line with square markers represent transitions from Th0 to Th1 and Th0 to Th2, respectively. The driver node is T-bet when Th0 is steered to the Th1 phenotype (Th0$\rightarrow$Th1) and GATA-3 when Th0 is driven to the Th2 phenotype (Th0$\rightarrow$ Th2)
Figure 6. Diagram of the cancer network. This network is redrawn from the reference ^{[33]}, including 32 nodes (genes) and 111 edges (66 activation interactions and 45 repression interactions). The network mainly includes four types of marker genes: apoptosis marker genes (green rectangles), cancer marker genes (red rectangles), tumor repressor genes (light blue rectangles) and other genes (blue rectangles).The red arrows represent activation and the green short bars represent repression
Figure 7. Domain control of the cancer network transitions from the cancer state to the apoptosis state and normal state. The representative node state trajectories $x(t)$ (A-D), control function $u(t)$ (E) and control error $e(t)$ (F) are depicted when AKT and RB are controlled, respectively. The symbols $x_4$, $x_6$, $x_{10}$ and $x_{25}$ indicate the activities of PTEN, RB, AKT and NF-$\kappa$B, respectively. The black dashed lines with star markers indicate the cancer state, i.e., the undesired state with no control. The blue solid lines with squares and red solid lines with filled circles represent transitions from cancer (C) to apoptosis (A) (denoted by C$\rightarrow$A) and cancer (C) to normal (N) (denoted by C$\rightarrow$N), respectively. The driver node is AKT when cancer is steered to the apoptosis and RB when cancer is driven to the normal state
Figure 8. Domain control strategies for phenotype transitions of the cancer network. The practically required driver nodes (genes) for realizing domain controllability of this network were presented. For example, the normal and apoptosis states can be induced from the cancer state by any one of the two genes (RB and CDK2) and any one of the three genes (AKT, PTEN and NF$\kappa$B), respectively
Figure A3. A schematic diagram of the relationships between the controlled variables (nodes) and the remaining variables. We relabel the indexes of variables of the system such that the control set $C= \{1, ..., m\}$. Then the ($m+1$)-th equation only contains the variables in the control set $C$. The ($m+2$)-th equation only contains the variables in the union of the control set $C$ and ($m+1$)-th node. By this analogy, the ($n$)-th equation contains the variables in the set $C\bigcup \{m+1, \cdots, n-1\}$
Figure C2. Domain control of the EMT core network transition from the epithelial state to the mesenchymal state by the theoretically required driver nodes. The node state trajectories $x(t)$ (A-F), control function $u(t)$ (G) and control error $e(t)$ (H) are depicted when the theoretically required driver nodes (i.e., miR-200 and miR-203) are controlled. The symbols $x_1, x_2, x_3, x_4, x_5$ and $x_6$ indicate the activities of CDH1, SNAI1, ZEB1, ZEB2, miR-200 and miR-203, respectively. $u_5$ and $u_6$ indicate the control functions for miR-200 and miR-203, respectively. The black dashed and blue solid lines represent the initial epithelial state (i.e., the undesired state with no control) and desired mesenchymal state, respectively
Figure C3. Effect of the parameter $\lambda$ in the designed control function on the domain control of the EMT core network. The node state trajectories $x(t)$ (A-F), control function $u(t)$ (G) and control error $e(t)$ (H) are depicted when the practically required driver node, namely miR-203, is controlled. The symbols $x_1, x_2, x_3, x_4, x_5$ and $x_6$ indicate the activities of CDH1, SNAI1, ZEB1, ZEB2, miR-200 and miR-203, respectively. The black dashed and colored solid lines represent the initial epithelial state (i.e., the undesired state with no control) and desired mesenchymal state, respectively. Here, we take the EMT core network as an example to show how the parameter $\lambda$ in our designed control function influences the control process. As shown in (H), to a certain extent, a higher absolute value of λ indicates a less time the system takes to evolve to the desirable attractor. However, there is no significant influence when the absolute value of $\lambda$ is larger than 1
Figure C5. Domain control of the T helper differentiation cellular network transition from Th1 to Th2. The representative node state trajectories $x(t)$ (A-K), control function $u(t)$ (L) and control error $e(t)$ (M) are depicted. The symbols $x_1, x_4, x_5, x_6, x_7, x_{12}, x_{13}, x_{17}, x_{19}, x_{21}$ and $x_{22}$ indicate the activities of GATA-3, IFN-$\gamma$, IFN-$\gamma$R, IL-10, IL-10R, IL-4, IL-4R, SOCS1, STAT3, STAT6 and T-bet, respectively. The dashed black line indicates the Th1 phenotype, i.e. the initial state with no control. The solid blue line represents the Th2 phenotype. The driver node is GATA-3 when Th1 is driven to the Th2 phenotype
Figure C7. Domain control of the cancer network transitions from the normal state to the apoptosis state and cancer state. The representative node state trajectories $x(t)$ (A-L), control function $u(t)$ (M) and control error $e(t)$ (N) are depicted when AKT ($u_{N\rightarrow A}$) and RB ($u_{N\rightarrow C}$) are controlled, respectively. The symbols $x_4, x_5, x_6, x_{10}, x_{12}, x_{13}, x_{14}, x_{16}, x_{18}, x_{21}, x_{22}$ and $x_{25}$ indicate the activities of PTEN, CDH1, RB, AKT, VEGF, HGF, HIF1, MDM2, CDK4, Caspase, BAX and NF$\kappa$B, respectively. The black dashed lines with star markers indicate the normal state, i.e., the initial state with no control. The red solid lines with filled circles and blue solid lines with squares represent transitions from normal (N) to apoptosis (A) (denoted by N$\rightarrow$A) and normal (A) to cancer (C) (denoted by N$\rightarrow$C), respectively. The driver node is AKT when normal is steered to the apoptosis and RB when normal is driven to the cancer state
Figure C8. Domain control of the cancer network transitions from the apoptosis state to the normal state and cancer state. The black dashed lines with star markers indicate the apoptosis state, i.e., the undesired state with no control. The red solid lines with filled circles and blue solid lines with squares represent transitions from the apoptosis state (A) to normal state (N) (denoted by A$\rightarrow$N) and apoptosis state (A) to cancer (C) (denoted by A$\rightarrow$C), respectively. The driver node is NF$\kappa$B when the apoptosis state is steered to the normal state and AKT when apoptosis state is driven to the cancer state
Figure C9. Robustness analysis of the control effectiveness against parameter perturbation. The evolution of the control function $u(t)$ and control error $e(t)$ are depicted when miR-203 is controlled. Each curve represents one simulation for one perturbation of the parameter values. In our numerical simulations, we randomly generate $100$ parameter sets in which every parameter of the system is perturbed within a range of $\pm10\%$. The initial state is the epithelial state, which is set to be $(1, 0, 0, 0, 1, 1)^T$ in all the numerical experiments. We zoom into the curves in a small window
Table 1. The attractors of the EMT network. The two observed attractors A and B correspond to the epithelial state and mesenchymal state ^{[41]}, respectively.
A | B | A | B | ||
$x_1$ (CDH1)} | 1.7924 | 0.1111 | $x_4$ (ZEB2)} | 0.0170 | 1.8169 |
$x_2$ (SNAI1)} | 0.3359 | 3.2229 | $x_5$ (miR-200)} | 1.7924 | 0.1111 |
$x_3$ (ZEB1)} | 0.0522 | 1.8224 | $x_6$ (miR-203)} | 2.9873 | 0.1851 |
Table 2. The positively invariant sets of the mathematical models of the three disease-related networks
Networks | Positively invariant sets |
EMT | $\left\{ ({{x}_{1}}, \cdots, {{x}_{6}})\in {{\mathbb{R}}^{6}}\left| 0 < {{x}_{i}} < \frac{1}{{{d}_{i}}}, i=1, \cdots, 6 \right. \right\}$ |
T helper | $\left\{x\in {{\mathbb{R}}^{23}}\left| \begin{aligned} & 0\le {{x}_{5}} < \frac{1}{{{d}_{4}}{{d}_{5}}}, 0\le {{x}_{6}} < \frac{1}{{{d}_{1}}{{d}_{6}}}, 0\le {{x}_{7}} < \frac{1}{{{d}_{1}}{{d}_{6}}{{d}_{7}}}, 0\le {{x}_{14}} < \frac{1}{{{d}_{11}}{{d}_{14}}}, \\ & 0\le {{x}_{17}} < ((\frac{1}{{{d}_{3}}}+\frac{1}{{{d}_{15}}})\frac{1}{{{d}_{18}}}+\frac{1}{{{d}_{22}}})\frac{1}{{{d}_{17}}}, 0\le {{x}_{18}} < (\frac{1}{{{d}_{3}}}+\frac{1}{{{d}_{15}}})\frac{1}{{{d}_{18}}}, \\ & 0\le {{x}_{19}} < \frac{1}{{{d}_{1}}{{d}_{6}}{{d}_{7}}{{d}_{19}}}, 0\le {{x}_{20}} < \frac{1}{{{d}_{9}}{{d}_{20}}}, 0\le {{x}_{21}} < \frac{1}{{{d}_{13}}{{d}_{21}}}, \\ & 0\le {{x}_{i}} < \frac{1}{{{d}_{i}}}, 0\le {{x}_{j}} < 1, i=1, 3, 4, 9, 11, 12, 13, 15, 16, j=8, 10, 23 \\ \end{aligned} \right. \right\}$ |
Cancer | $\left\{ ({{x}_{1}}, \cdots, {{x}_{32}})\in {{\mathbb{R}}^{32}}\left| 0\le {{x}_{i}}\le 1, i=1, \cdots, 32 \right. \right\}$ |
Table 3. The theoretically required driver nodes and designed control functions for realizing domain control of the three disease-related networks. $x_{2, i}^{*}$ is the $i$-th component of the state vector of the desired attractor
Networks | Driver nodes | Control functions ($u(t)$) |
EMT | miR-200 and miR-203 | $\left[\begin{matrix} -\left(\frac{1}{1+{{K}_{52}}x_{2}^{2}+{{K}_{53}}x_{3}^{2}+{{K}_{54}}x_{4}^{2}}-{{d}_{5}}{{x}_{5}} \right)+\lambda \left({{x}_{5}}-x_{2, 5}^{*} \right) \\ -\left(\frac{1}{1+{{K}_{62}}x_{2}^{2}+{{K}_{63}}x_{3}^{2}+{{K}_{64}}x_{4}^{2}}-{{d}_{6}}{{x}_{6}} \right)+\lambda \left({{x}_{6}}-x_{2, 6}^{*} \right) \\ \end{matrix} \right]$ |
T helper | GATA-3 and STAT1 | $\left[\begin{matrix} -\left(\frac{{{k}_{1-1}}{{x}_{1}}+{{k}_{1-21}}{{x}_{21}}}{1+{{k}_{1-1}}{{x}_{1}}+{{k}_{1-21}}{{x}_{21}}+{{k}_{1-22}}{{x}_{22}}}-{{d}_{1}}{{x}_{1}} \right)+\lambda \left({{x}_{1}}-x_{2, 1}^{*} \right) \\ -\left({{k}_{18-3}}{{x}_{3}}+{{k}_{18-15}}{{x}_{15}}-{{d}_{18}}{{x}_{18}} \right)+\lambda \left({{x}_{18}}-x_{2, 18}^{*} \right) \\ \end{matrix} \right]$ |
Cancer | P53, RB, AKT, EGFR, HIF1, CDK2, CDK1, BCL2 and NF$\kappa$B | $\left[\begin{matrix} -\left(\frac{a\left({{s}_{1}}{{x}_{1}}^{n}+{{s}_{9}}{{x}_{9}}^{n}+{{s}_{14}}{{x}_{14}}^{n} \right)}{3{{S}^{n}}+{{s}_{1}}{{x}_{1}}^{n}+{{s}_{9}}{{x}_{9}}^{n}+{{s}_{14}}{{x}_{14}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+1/2{{s}_{16}}{{x}_{16}}^{n}+1/2{{s}_{31}}{{x}_{31}}^{n}}-k{{x}_{2}} \right)+\lambda \left({{x}_{2}}-x_{2, 2}^{*} \right) \\ -\left(\frac{2b{{S}^{n}}}{{{S}^{n}}+1/2{{s}_{17}}{{x}_{17}}^{n}+1/2{{s}_{18}}{{x}_{18}}^{n}}-k{{x}_{6}} \right)+\lambda \left({{x}_{6}}-x_{2, 6}^{*} \right) \\ -\left(\frac{a\left({{s}_{11}}{{x}_{11}}^{n}+{{s}_{12}}{{x}_{12}}^{n}+{{s}_{13}}{{x}_{13}}^{n} \right)/3}{{{S}^{n}}+1/3{{s}_{11}}{{x}_{11}}^{n}+1/3{{s}_{12}}{{x}_{12}}^{n}+1/3{{s}_{13}}{{x}_{13}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{4}}{{x}_{4}}^{n}}-k{{x}_{10}} \right)+\lambda \left({{x}_{10}}-x_{2, 10}^{*} \right) \\ -\left(\frac{a\left({{s}_{9}}{{x}_{9}}^{n}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{11}}{{x}_{11}}^{n}+{{s}_{13}}{{x}_{13}}^{n}+{{s}_{14}}{{x}_{14}}^{n}+{{s}_{15}}{{x}_{15}}^{n} \right)}{6{{S}^{n}}+{{s}_{9}}{{x}_{9}}^{n}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{11}}{{x}_{11}}^{n}+{{s}_{13}}{{x}_{13}}^{n}+{{s}_{14}}{{x}_{14}}^{n}+{{s}_{15}}{{x}_{15}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{2}}{{x}_{2}}^{n}}-k{{x}_{12}} \right)+\lambda \left({{x}_{12}}-x_{2, 12}^{*} \right) \\ -\left(\frac{a\left({{s}_{10}}{{x}_{10}}^{n}+{{s}_{13}}{{x}_{13}}^{n}+{{s}_{25}}{{x}_{25}}^{n}+{{s}_{26}}{{x}_{26}}^{n}+{{s}_{28}}{{x}_{28}}^{n} \right)}{5{{S}^{n}}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{13}}{{x}_{13}}^{n}+{{s}_{25}}{{x}_{25}}^{n}+{{s}_{26}}{{x}_{26}}^{n}+{{s}_{28}}{{x}_{28}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{2}}{{x}_{2}}^{n}}-k{{x}_{14}} \right)+\lambda \left({{x}_{14}}-x_{2, 14}^{*} \right) \\ -\left(\frac{a\left({{s}_{9}}{{x}_{9}}^{n}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{20}}{{x}_{20}}^{n} \right)}{3{{S}^{n}}+{{s}_{9}}{{x}_{9}}^{n}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{20}}{{x}_{20}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+1/2{{s}_{3}}{{x}_{3}}^{n}+1/2{{s}_{6}}{{x}_{6}}^{n}}-k{{x}_{17}} \right)+\lambda \left({{x}_{17}}-x_{2, 17}^{*} \right) \\ -\left(\frac{2b{{S}^{n}}}{{{S}^{n}}+1/3{{s}_{3}}{{x}_{3}}^{n}+1/3{{s}_{5}}{{x}_{5}}^{n}+1/3{{s}_{30}}{{x}_{30}}^{n}}-k{{x}_{19}} \right)+\lambda \left({{x}_{19}}-x_{2, 19}^{*} \right) \\ -\left(\frac{a\left({{s}_{22}}{{x}_{22}}^{n}+{{s}_{23}}{{x}_{23}}^{n} \right)/2}{{{S}^{n}}+1/2{{s}_{22}}{{x}_{22}}^{n}+1/2{{s}_{23}}{{x}_{23}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+1/2{{s}_{10}}{{x}_{10}}^{n}+1/2{{s}_{24}}{{x}_{24}}^{n}}-k{{x}_{21}} \right)+\lambda \left({{x}_{21}}-x_{2, 21}^{*} \right) \\ -\left(\frac{a\left({{s}_{10}}{{x}_{10}}^{n}+{{s}_{13}}{{x}_{13}}^{n} \right)/2}{{{S}^{n}}+1/2{{s}_{10}}{{x}_{10}}^{n}+1/2{{s}_{13}}{{x}_{13}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{4}}{{x}_{4}}^{n}}-k{{x}_{25}} \right)+\lambda \left({{x}_{25}}-x_{2, 25}^{*} \right) \\ \end{matrix} \right]$ |
Table 4. An example of the practically required driver nodes and designed control functions for realizing domain control of the three disease-related networks. E and M represent the epithelial state and mesenchymal state, respectively. C, A and N represent the cancer, apoptosis and normal states, respectively. $x_{2, i}^{*}$ is the $i$-th component of the state vector of the desired attractor
Transitions | Drivers | Control functions ($u(t)$) |
E$\leftrightarrow$M | miR-203 | ${-\left(\frac{1}{1+{{K}_{62}}x_{2}^{2}+{{K}_{63}}x_{3}^{2}+{{K}_{64}}x_{4}^{2}}-{{d}_{6}}{{x}_{6}} \right)+\lambda \left({{x}_{6}}-x_{2, 6}^{*} \right)}$ |
Th0$\leftrightarrow$Th1 | T-bet | $ {-\left(\frac{{{k}_{22-18}}{{x}_{18}}+{{k}_{22-22}}{{x}_{22}}}{1+{{k}_{22-1}}{{x}_{1}}+{{k}_{22-22}}{{x}_{22}}+{{k}_{22-18}}{{x}_{18}}}-{{d}_{22}}{{x}_{22}} \right)+\lambda \left({{x}_{22}}-x_{2, 22}^{*} \right)}$ |
Th0$\leftrightarrow$Th2 | GATA-3 | $ {-\left(\frac{{{k}_{1-1}}{{x}_{1}}+{{k}_{1-21}}{{x}_{21}}}{1+{{k}_{1-1}}{{x}_{1}}+{{k}_{1-21}}{{x}_{21}}+{{k}_{1-22}}{{x}_{22}}}-{{d}_{1}}{{x}_{1}} \right)+\lambda \left({{x}_{1}}-x_{2, 1}^{*} \right)}$ |
C$\leftrightarrow$A | AKT | ${-\left(\frac{a\left({{s}_{11}}{{x}_{11}}^{n}+{{s}_{12}}{{x}_{12}}^{n}+{{s}_{13}}{{x}_{13}}^{n} \right)}{3{{S}^{n}}+{{s}_{11}}{{x}_{11}}^{n}+{{s}_{12}}{{x}_{12}}^{n}+{{s}_{13}}{{x}_{13}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{4}}{{x}_{4}}^{n}}-k{{x}_{10}} \right)+\lambda \left({{x}_{10}}-x_{2, 10}^{*} \right)}$ |
C$\leftrightarrow$N | RB | ${-\left(\frac{4b{{S}^{n}}}{2{{S}^{n}}+{{s}_{17}}{{x}_{17}}^{n}+{{s}_{18}}{{x}_{18}}^{n}}-k{{x}_{6}} \right)+\lambda \left({{x}_{6}}-x_{2, 6}^{*} \right)}$ |
A$\leftrightarrow$N | NF$\kappa$B | ${-\left(\frac{a\left({{s}_{10}}{{x}_{10}}^{n}+{{s}_{13}}{{x}_{13}}^{n} \right)}{2{{S}^{n}}+{{s}_{10}}{{x}_{10}}^{n}+{{s}_{13}}{{x}_{13}}^{n}}+\frac{b{{S}^{n}}}{{{S}^{n}}+{{s}_{4}}{{x}_{4}}^{n}}-k{{x}_{25}} \right)+\lambda \left({{x}_{25}}-x_{2, 25}^{*} \right)}$ |
Table 5. The attractors of the Th differentiation cellular network. The three observed attractors (A, B and C) correspond to the Th0, Th1 and Th2 phenotypes ^{[17]}, respectively
A | B | C | A | B | C | ||
$x_{1}$ (GATA-3) | 0.0000 | 0.0000 | 2.0238 | $x_{13}$ (IL-4R) | 0.0000 | 0.0000 | 1.3358 |
$x_{2}$ (IFN-$\beta$) | 0.0000 | 0.0000 | 0.0000 | $x_{14}$ (IRAK) | 0.0000 | 0.0000 | 0.0000 |
$x_{3}$ (IFN-$\beta$R) | 0.0000 | 0.0000 | 0.0000 | $x_{15}$ (JAK1) | 0.0000 | 0.4406 | 0.0000 |
$x_{4}$ (IFN-$\gamma$) | 0.0000 | 2.1012 | 0.0000 | $x_{16}$ (NFAT) | 0.0000 | 0.0000 | 0.0000 |
$x_{5}$ (IFN-$\gamma$R) | 0.0000 | 2.1012 | 0.0000 | $x_{17}$ (SOCS1) | 0.0000 | 2.1459 | 0.0000 |
$x_{6}$ (IL-10) | 0.0000 | 0.0000 | 2.0238 | $x_{18}$ (STAT1) | 0.0000 | 0.4406 | 0.0000 |
$x_{7}$ (IL-10R) | 0.0000 | 0.0000 | 2.0238 | $x_{19}$ (STAT3) | 0.0000 | 0.0000 | 2.0238 |
$x_{8}$ (IL-12) | 0.0000 | 0.0000 | 0.0000 | $x_{20}$ (STAT4) | 0.0000 | 0.0000 | 0.0000 |
$x_{9}$ (IL-12R) | 0.0000 | 0.0000 | 0.0000 | $x_{21}$ (STAT6) | 0.0000 | 0.0000 | 2.2264 |
$x_{10}$ (IL-18) | 0.0000 | 0.0000 | 0.0000 | $x_{22}$ (T-bet) | 0.0000 | 1.7053 | 0.0000 |
$x_{11}$ (IL-18R) | 0.0000 | 0.0000 | 0.0000 | $x_{23}$ (TCR) | 0.0000 | 0.0000 | 0.0000 |
$x_{12}$ (IL-4) | 0.0000 | 0.0000 | 2.0094 |
Table 6. The attractors of the cancer network. The three observed attractors (A, B and C) correspond to the apoptosis, normal and cancer states ^{[33]}, respectively.
A | B | C | A | B | C | ||
$x_1$ (ATM) | 0.4165 | 0.4277 | 0.4712 | $x_{17}$ (CDK2) | 0.1336 | 0.4613 | 0.8022 |
$x_2$ (P53) | 0.4668 | 0.4642 | 0.4545 | $x_{18}$ (CDK4) | 0.1342 | 0.4273 | 0.8620 |
$x_3$ (P21) | 0.5700 | 0.4511 | 0.4378 | $x_{19}$ (CDK1) | 0.5788 | 0.4853 | 0.5550 |
$x_4$ (PTEN) | 0.7438 | 0.2973 | 0.2836 | $x_{20}$ (E2F1) | 0.1922 | 0.2620 | 0.3441 |
$x_5$ (CDH1) | 0.3215 | 0.6263 | 0.5298 | $x_{21}$ (Caspase) | 0.8766 | 0.0688 | 0.0621 |
$x_6$ (RB) | 0.9970 | 0.7105 | 0.1921 | $x_{22}$ (BAX) | 0.7159 | 0.2641 | 0.2498 |
$x_7$ (ARF) | 0.2756 | 0.2645 | 0.3087 | $x_{23}$ (BAD) | 0.8486 | 0.1024 | 0.0923 |
$x_8$ (AR) | 0.4134 | 0.2439 | 0.1955 | $x_{24}$ (BCL2) | 0.1740 | 0.7533 | 0.7705 |
$x_9$ (MYC) | 0.6647 | 0.4760 | 0.4758 | $x_{25}$ (NF$\kappa$B) | 0.1433 | 0.8853 | 0.9007 |
$x_{10}$ (AKT) | 0.3044 | 0.8058 | 0.8294 | $x_{26}$ (RAS) | 0.4089 | 0.5158 | 0.5427 |
$x_{11}$ (EGFR) | 0.5262 | 0.4606 | 0.4636 | $x_{27}$ (TGF$\alpha$) | 0.0000 | 0.0000 | 0.0000 |
$x_{12}$ (VEGF) | 0.4145 | 0.6239 | 0.6464 | $x_{28}$ (TNF$\alpha$) | 0.0000 | 0.0000 | 0.0000 |
$x_{13}$ (HGF) | 0.1484 | 0.5908 | 0.6214 | $x_{29}$ (TGF$\beta$) | 0.0806 | 0.6772 | 0.7016 |
$x_{14}$ (HIF1) | 0.2998 | 0.6632 | 0.6823 | $x_{30}$ (Wee1) | 0.6282 | 0.4550 | 0.5882 |
$x_{15}$ (hTERT) | 0.3765 | 0.4680 | 0.4702 | $x_{31}$ (MdmX) | 0.6915 | 0.8148 | 0.6413 |
$x_{16}$ (MDM2) | 0.2471 | 0.4911 | 0.7550 | $x_{32}$ (Wip1) | 0.4933 | 0.4877 | 0.4666 |
[1] |
P. Ao, D. Galas, L. Hood and X. M. Zhu, Cancer as robust intrinsic state of endogenous molecular-cellular network shaped by evolution, Med. Hypotheses, 70 (2008), 678-684. |
[2] |
R. P. Araujo, L. A. Liotta and E. F. Petricoin, Proteins, drug targets and the mechanisms they control: The simple truth about complex networks, Nat. Rev. Drug Discov., 6 (2007), 871-880. |
[3] |
K. Axelrod, A. Sanchez and J. Gore, Phenotypic states become increasingly sensitive to perturbations near a bifurcation in a synthetic gene network, eLife, 4 (2015), e07935. |
[4] |
S. Balint, E. Kaslik, A. M. Balint and A. Grigis, Methods for determination and approximation of the domain of attraction in the case of autonomous discrete dynamical systems, Adv. Differ. Equ., 2006 (2006), Art. 23939, 1-15. |
[5] |
Y. Bar-Yam, D. Harmon and B. de Bivort, Attractors and democratic dynamics, Science, 323 (2009), 1016-1017. |
[6] |
A.-L. Barabási, N. Gulbahce and J. Loscalzo, Network medicine: A network-based approach to human disease, Nat. Rev. Genet., 12 (2011), 56-68. |
[7] | R. G. Bartle, The Elements of Integration and Lebesgue Measure, (John Wiley & Sons), 2011. |
[8] |
B. Barzel and A.-L. Barabási, Universality in network dynamics, Nat. Phys., 9 (2013), 673-681. |
[9] |
Y. Ben-Neriah and M. Karin, Inflammation meets cancer, with nf-κb as the matchmaker, Nat.Immunol., 12 (2011), 715-723. |
[10] | G. Chen and X. Yu, Chaos Control: Theory and Applications, Springer-Verlag Berlin Heidelberg, Berlin, 2003. |
[11] | Y. -Z. Chen, L. Wang, W. Wang and Y. -C. Lai, The paradox of controlling complex networks: Control inputs versus energy requirement, preprint, arXiv: 1509.03196. |
[12] |
S. P. Cornelius, W. L. Kath and A. E. Motter, Realistic control of network dynamics, Nat. Commun., 4 (2013), 1942. |
[13] |
P. Creixell, E. M. Schoof, J. T. Erler and R. Linding, Navigating cancer network attractors for tumor-specific therapy, Nat. Biotechnol., 30 (2012), 842-848. |
[14] |
P. Csermely, T. Korcsmáros, H. J. Kiss, G. London and R. Nussinov, Structure and dynamics of molecular networks: A novel paradigm of drug discovery: A comprehensive review, Pharmacol & Therapeut, 138 (2013), 333-408. |
[15] |
B. De Craene and G. Berx, Regulatory networks defining EMT during cancer initiation and progression, Nat. Rev. Cancer, 13 (2013), 97-110. |
[16] |
H. De Jong, Modeling and simulation of genetic regulatory systems: A literature review, J. Comput. Biol., 9 (2002), 67-103. |
[17] |
A. Di Cara, A. Garg, G. De Micheli, I. Xenarios and L. Mendoza, Dynamic simulation of regulatory networks using SQUAD, BMC Bioinformatics, 8 (2007), p462. |
[18] |
B. Fiedler, A. Mochizuki, G. Kurosawa and D. Saito, Dynamics and control at feedback vertex sets. Ⅰ: informative and determining nodes in regulatory networks, J. Dyn. Differ. Equ., 25 (2013), 563-604. |
[19] |
T. S. Gardner, C. R. Cantor and J. J. Collins, Construction of a genetic toggle switch in Escherichia coli, Nature, 403 (2000), 339-342. |
[20] |
J. Gao, Y.-Y. Liu, R. M. D'Souza and A.-L. Barabási, Target control of complex networks, Nat. Commun., 5 (2014), p5415. |
[21] |
B. T. Hennessy, D. L. Smith, P. T. Ram, Y. Lu and G. B. Mills, Exploiting the PI3K/AKT pathway for cancer drug discovery, Nat. Rev. Drug Discov., 4 (2005), 988-1004. |
[22] |
S. Huang, I. Ernberg and S. Kauffman, Cancer attractors: a systems view of tumors from a gene network dynamics and developmental perspective, Semin. Cell Dev. Biol., 20 (2009), 869-876. |
[23] |
E. S. Hwang, S. J. Szabo, P. L. Schwartzberg and L. H. Glimcher, T helper cell fate specified by kinase-mediated interaction of T-bet with GATA-3, Science, 307 (2005), 430-433. |
[24] |
R. G. Jenner, M. J. Townsend, I. Jackson, K. Sun, R. D. Bouwman, R. A. Young, L. H. Glimcher and G. M. Lord, The transcription factors T-bet and GATA-3 control alternative pathways of T-cell differentiation through a shared set of target genes, Proc. Natl. Acad. Sci. U. S. A., 106 (2009), 17876-17881. |
[25] |
S. Jin, Y. Li, R. Pan and X. Zou, Characterizing and controlling the inflammatory network during influenza a virus infection, Sci. Rep., 4 (2014), p3799. |
[26] |
S. Jin, L. Niu, G. Wang and X. Zou, Mathematical modeling and nonlinear dynamical analysis of cell growth in response to antibiotics, Int. J. Bifurcat. Chaos, 25 (2015), 1540007, 12pp. |
[27] |
J. D. Jordan, E. M. Landau and R. Iyengar, Signaling networks: The origins of cellular multitasking, Cell, 103 (2000), 193-200. |
[28] |
R. Kalluri and R. A. Weinberg, The basics of epithelial-mesenchymal transition, J. Clin. Invest., 119 (2009), 1420-1428. |
[29] |
S. Karl and T. Dandekar, Convergence behaviour and control in non-linear biological networks, Sci. Rep., 5 (2015), p9746. |
[30] | H. K. Khalil, Nonlinear Systems, 3rd edition, Prentice Hall, New Jersey, 2002. |
[31] |
Y. C. Lai, Controlling complex, non-linear dynamical networks, National Science Review, 1 (2014), 339-341. |
[32] |
S. Lamouille, J. Xu and R. Derynck, Molecular mechanisms of epithelial-mesenchymal transition, Nat. Rev. Mol. Cell Biol., 15 (2014), 178-196. |
[33] |
C. Li and J. Wang, Quantifying the underlying landscape and paths of cancer, J. R. Soc. Interface, 11 (2014), 20140774. |
[34] |
Y. Li, S. Jin, L. Lei, Z. Pan and X. Zou, Deciphering deterioration mechanisms of complex diseases based on the construction of dynamic networks and systems analysis, Sci. Rep., 5 (2015), p9283. |
[35] |
Y. Li, M. Yi and X. Zou, The linear interplay of intrinsic and extrinsic noises ensures a high accuracy of cell fate selection in budding yeast, Sci. Rep., 4 (2014), p5764. |
[36] |
Y. Y. Liu and A. L. Barabasi, Control principles of complex systems, Rev. Mod. Phys., 88 (2016), p035006. |
[37] |
Y. Y. Liu, J. J. Slotine and A. L. Barabasi, Controllability of complex networks, Nature, 473 (2011), 167-173. |
[38] |
L. G. Matallana, A. M. Blanco and J. A. Bandoni, Estimation of domains of attraction: A global optimization approach, Math. Comput. Model., 52 (2010), 574-585. |
[39] |
L. Mendoza, A network model for the control of the differentiation process in Th cells, Biosystems, 84 (2006), 101-114. |
[40] |
A. Mochizuki, B. Fiedler, G. Kurosawa and D. Saito, Dynamics and control at feedback vertex sets. Ⅱ: A faithful monitor to determine the diversity of molecular activities in regulatory networks, J. Theor. Biol., 335 (2013), 130-146. |
[41] |
M. Moes, A. Le Béchec, I. Crespo, C. Laurini, A. Halavatyi, G. Vetter, A. Del Sol and E. Friederich, A novel network integrating a miRNA-203/SNAI1 feedback loop which regulates epithelial to mesenchymal transition, PloS One, 7 (2012), e35440. |
[42] |
F.-J. Müller and A. Schuppert, Few inputs can reprogram biological networks, Nature, 478 (2011), E4-E4. |
[43] |
T. Nepusz and T. Vicsek, Controlling edge dynamics in complex networks, Nat. Phys., 8 (2012), 568-573. |
[44] |
J. A. Papin, T. Hunter, B. O. Palsson and S. Subramaniam, Reconstruction of cellular signalling networks and analysis of their properties, Nat. Rev. Mol. Cell Biol., 6 (2005), 99-111. |
[45] |
J. Pei, N. Yin, X. Ma and L. Lai, Systems biology brings new dimensions for structure-based drug design, J. Am. Chem. Soc., 136 (2014), 11556-11565. |
[46] |
M. Pósfai and P. Hövel, Structural controllability of temporal networks, New J. Phys., 16 (2014), 123055. |
[47] |
J. Ruths and D. Ruths, Control profiles of complex networks, Science, 343 (2014), 1373-1376. |
[48] | J. -J. E. Slotine and W. Li, Applied Nonlinear Control, Prentice-Hall, New Jersey, 1991. |
[49] |
M. S. Song, L. Salmena and P. P. Pandolfi, The functions and regulation of the pten tumour suppressor, Nat. Rev. Mol. Cell Biol., 13 (2012), 283-296. |
[50] |
F. Sorrentino, M. di Bernardo, F. Garofalo and G. R. Chen, Controllability of complex networks via pinning, Phys. Rev. E, 75 (2007), 046103. |
[51] | G. Strang, Calculus, Wellesley-Cambridge Press, Massachusetts, 1991. |
[52] |
J. Tan and X. Zou, Complex dynamical analysis of a coupled network from innate immune responses, Int. J. Bifurcat. Chaos, 23 (2013), 1350180, 26pp. |
[53] |
J. Tan and X. Zou, Optimal control strategy for abnormal innate immune response, Comput. Math. Method M., 2015 (2015), Art. ID 386235, 16 pp. |
[54] |
A. Vinayagam, T. E. Gibson, H.-J. Lee, B. Yilmazel, C. Roesel, Y. Hu, Y. Kwon, A. Sharma, Y.-Y. Liu, N. Perrimon and A.-L. Barabási, Controllability analysis of the directed human protein inteorkraction netw identifies disease genes and drug targets, Proc. Natl. Acad. Sci. U. S. A., 113 (2016), 4976-4981. |
[55] |
D. Wang, S. Jin, F. X. Wu and X. Zou, Estimation of control energy and control strategies for complex networks, Adv. Complex Syst., 18 (2015), 1550018, 23pp. |
[56] |
D. Wang, S. Jin and X. Zou, Crosstalk between pathways enhances the controllability of signalling networks, IET Syst. Biol., 10 (2016), 2-9. |
[57] |
L. Z. Wang, R. Q. Su, Z. G. Huang, X. Wang, W. X. Wang, C. Grebogi and Y. C. Lai, A geometrical approach to control and controllability of nonlinear dynamical networks, Nat. Commun., 7 (2016), p11323. |
[58] |
Y. Wang, J. Tan, F. Sadre-Marandi, J. Liu and X. Zou, Mathematical modeling for intracellular transport and binding of HIV-1 gag proteins, Math. Biosci., 262 (2015), 198-205. |
[59] | R. Weinberg, The Biology of Cancer, 2^{nd} edition, Garland Science, New York, 2013. |
[60] |
U. Wellner, J. Schubert, U. C. Burk, O. Schmalhofer, F. Zhu, A. Sonntag, B. Waldvogel, C. Vannier, D. Darling, A. zur Hausen et al., The EMT-activator ZEB1 promotes tumorigenicity by repressing stemness-inhibiting micrornas, Nat. Cell Biol., 11 (2009), 1487-1495. |
[61] |
D. K. Wells, W. L. Kath and A. E. Motter, Control of stochastic and induced switching in biophysical networks, Phys. Rev. X, 5 (2015), 031036. |
[62] |
A. J. Whalen, S. N. Brennan, T. D. Sauer and S. J. Schiff, Observability and controllability of nonlinear networks: The role of symmetry, Phys. Rev. X, 5 (2015), 011005. |
[63] |
F.-X. Wu, L. Wu, J. Wang, J. Liu and L. Chen, Transittability of complex networks and its applications to regulatory biomolecular networks, Sci. Rep., 4 (2014), p4819. |
[64] |
S. Wuchty, Controllability in protein interaction networks, Proc. Natl. Acad. Sci. U. S. A., 111 (2014), 7156-7160. |
[65] |
G. Yan, G. Tsekenis, B. Barzel, J.-J. Slotine, Y.-Y. Liu and A.-L. Barabási, Spectrum of controlling and observing complex networks, Nat. Phys., 11 (2015), 779-786. |
[66] |
Z. Yuan, C. Zhao, Z. Di, W.-X. Wang and Y.-C. Lai, Exact controllability of complex networks, Nat. Commun., 4 (2013), p2447. |
[67] | J. G. Zañudo, G. Yang and R. Albert, Structure-based control of complex networks with nonlinear dynamics, arXiv: 1605.08415v2. |
[68] |
J. Zhang, Z. Yuan, H. X. Li and T. Zhou, Architecture-dependent robustness and bistability in a class of genetic circuits, Biophys. J., 99 (2010), 1034-1042. |
[69] |
N. Zhong, Computational unsolvability of domains of attraction of nonlinear systems, Proc. Amer. Math. Soc., 137 (2009), 2773-2783. |