1: \documentclass[twocolumn,aps,showpacs,showkeys,floatfix]{revtex4}

2: \usepackage{amssymb,amsmath}

3: \usepackage[pdftex]{graphicx}

4: %\usepackage[T1]{fontenc}

5: %\usepackage[latin1]{inputenc}

6: 

7: \bibliographystyle{apsrev}

8: 

9: \renewcommand{\epsilon }{\varepsilon}

10: \renewcommand{\phi}{\varphi}

11: \renewcommand{\tilde}{\widetilde}

12: \renewcommand{\hat}{\widehat}

13: \renewcommand{\Im}{\mathrm{Im}}

14: \renewcommand{\Re}{\mathrm{Re}}

15: \renewcommand{\AA}{\mathcal{A}}

16: \newcommand{\BB}{\mathcal{B}}

17: \newcommand{\CC}{\mathcal{C}}

18: \newcommand{\DD}{\mathcal{D}}

19: \newcommand{\EE}{\mathcal{E}}

20: \newcommand{\FF}{\mathcal{F}}

21: \newcommand{\GG}{\mathcal{G}}

22: \newcommand{\HH}{\mathcal{H}}

23: \newcommand{\II}{\mathcal{I}}

24: \newcommand{\JJ}{\mathcal{J}}

25: \newcommand{\KK}{\mathcal{K}}

26: \newcommand{\LL}{\mathcal{L}}

27: \newcommand{\MM}{\mathcal{M}}

28: \newcommand{\NN}{\mathcal{N}}

29: \newcommand{\OO}{\mathcal{O}}

30: \newcommand{\PP}{\mathcal{P}}

31: \newcommand{\QQ}{\mathcal{Q}}

32: \newcommand{\RR}{\mathcal{R}}

33: \renewcommand{\SS}{\mathcal{S}}

34: \newcommand{\TT}{\mathcal{T}}

35: \newcommand{\UU}{\mathcal{U}}

36: \newcommand{\VV}{\mathcal{V}}

37: \newcommand{\WW}{\mathcal{W}}

38: \newcommand{\XX}{\mathcal{X}}

39: \newcommand{\YY}{\mathc/epfal{Y}}

40: \newcommand{\ZZ}{\mathcal{Z}}

41: %%plot symbols

42: \newcommand{\SA}{\circ}

43: \newcommand{\SB}{\bigtriangleup}

44: \newcommand{\SC}{+}

45: \newcommand{\SD}{\times}

46: \newcommand{\SE}{\diamond}

47: \newcommand{\SF}{\bigtriangledown}

48: 

49: \newcommand{\real}{\mathbb{R}}

50: \newcommand{\complex}{\mathbb{C}}

51: \newcommand{\integer}{\mathbb{Z}}

52: \renewcommand{\natural}{\mathbb{N}}

53: 

54: \newcommand{\HALF}{\frac{1}{2}}

55: \newcommand{\FOUR}{\frac{1}{4}}

56: 

57: %

58: \def\be{\begin{equation}}

59: \def\ee{\end{equation}}

60: \def\bq{\begin{eqnarray}}

61: \def\eq{\end{eqnarray}}

62: \def\beq{\begin{eqnarray}}

63: \def\eeq{\end{eqnarray}}

64: \def\ov{\overline}

65: \def\un{\underline}

66: %

67: 

68: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

69: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

70: 

71: \begin{document} 

72: 

73: %\title{Dynamical characteristics of perturbed biological networks}

74: 

75: \title{Representing perturbed dynamics in biological network models}

76: \author{Gautier Stoll$^{1,3}$}

77: \email{Gautier.stoll@curie.fr}

78: \author{Jacques Rougemont$^{3}$}

79: \email{jacques.rougemont@isb-sib.ch}

80: \author{Felix Naef$^{1,2,3}$}

81: \email{felix.naef@isrec.ch}

82: 

83: \affiliation{$^1$NCCR Molecular Oncology, ch. des Boveresses 155, 1066

84:   Epalinges, Switzerland} 

85: \affiliation{$^3$School of Life Sciences, ISREC, Ecole polytechnique

86:   F\'ed\'erale de Lausanne 1015 Lausanne, Switzerland} 

87: \affiliation{$^3$Swiss Institute of Bioinformatics, Quartier Sorge-Genopode, 1015

88:   Lausanne, Switzerland} 

89: 

90: \begin{abstract}

91: We study the dynamics of gene activities in relatively small size biological networks (up to a few tens of nodes), e.g. the 

92: activities of cell-cycle proteins during the mitotic cell-cycle progression.

93: Using the framework of deterministic discrete dynamical models, we characterize the dynamical modifications in response

94: to structural perturbations in the network connectivities.

95: In particular, we focus on how perturbations affect the set of fixed points and sizes of the basins of attraction.

96: Our approach uses two analytical measures: the basin entropy $H$ and the perturbation size $\Delta$, a quantity that reflects

97: the distance between the set of fixed points of the perturbed network to that of

98: the unperturbed network.

99: Applying our approach to the yeast-cell cycle network introduced by

100: Li \textit{et al.}

101: provides a low dimensional and informative fingerprint of network behavior under

102: large classes of perturbations. We identify

103: interactions that are crucial for proper network function, and also pinpoints

104: functionally redundant network connections. 

105: Selected perturbations exemplify the breadth of dynamical responses in this cell-cycle model.

106: \end{abstract}

107: 

108: \keywords{}

109: \pacs{}

110: \maketitle

111: 

112: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

114: 

115: \section{Introduction}  

116: 

117: Recent experimental developments in the fields of genomics, e.g. whole genome DNA sequencing or proteomics,

118: are opening possibilities for systems level studies in biology \cite{leibler_3,arkin_4,barabasi_3,alon_8}.

119: In particular, the notion that biological functions may rely on a large number of interconnected variables (for example genes) working in concert has stimulated general theoretical interest

120: about properties of biological networks \cite{barabasi_2}.

121: Studies of the statistical properties of large (typically thousands of nodes) biological networks have identified

122: a number of functional building block, termed network motifs, that occur more frequently than random \cite{alon_9}.

123: These findings support the idea that some systems are designed around a modular architecture, in which autonomous modules are wired

124: together to generate versatile biological functions \cite{leibler_3,barkai_9,alon_8,eckmann}.

125: While structural (or topological) properties are key for network characterization, functional properties are ultimately

126: encoded in dynamical, or time-dependent changes in the state variables of the nodes.

127: The sizes of systems that can be modeled dynamically

128: are typically much smaller (10-100 nodes). One common modeling approach, for example for the yeast cell-cycle \cite{tyson_novak_5},

129: is to simulate the nonlinear system of chemical rate equations describing the putative biochemical processes.

130: Modeling approaches have been applied to a number of systems,

131: including the cell-cycle \cite{tyson_novak_5,cross_3}, the lambda-phage switch in E. coli \cite{arkin2_4}.

132: Although these models provide a detailed description, this approach suffers from the caveat that most parameters are currently not accessible experimentally.

133: In addition, the number of parameters is typically about five per reaction,

134: resulting in a prohibitively large parameter space.

135: This last point makes it difficult to grasp the full solution space of the model. Recent approaches based on

136: sampling the parameter space in optimal regions have been developed \cite{brown_sethna_1}.

137: At the opposite end of model complexity, dynamical rules based on boolean state variables have been useful

138: for studying more global dynamical properties of topological classes of networks \cite{kadanoff,kauffman_1}.

139: In addition, boolean models have been successfully applied to the yeast cell-cycle \cite{tang_1, stoll} and the body patterning

140: in drosophila embryos \cite{albert_3,albert_4}.

141: 

142: In this study, we develop a systematic approach to describe how the dynamical

143: landscape of small (less than about 50 nodes) boolean networks is affected by perturbations in the network connectivity.

144: In particular, we consider the basin entropy $H$, a quantity that considers the size distribution of

145: the basins of attraction.

146: We complement entropy with a measure of distance between the stable

147: fixed points of a perturbed network and those in the unperturbed network. This combination

148: gives a low-dimensional and compact representation of the patterns induced by a

149: large number of perturbations. We illustrate our methods using the yeast cell-cycle network

150: introduced in \cite{tang_1}, and discuss examples of structural perturbations producing a range

151: of modified basins of attraction. 

152: 

153: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

154: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

155: 

156: \section{Definitions}  

157: Following \cite{tang_1} a network of $N$ nodes can be represented by a $N\times N$ adjacency matrix $A$, in which an activating link between node $i$ and node $j$ is 

158: represented by $A_{ij}=1$ and an inhibiting link by $A_{ij}=-1$.

159: The possibility of self-inhibitory (or activating links) $A_{ii}=\pm1$ is not excluded.

160: In the Boolean approximation, each node has two possible states,

161: so that the global state of all nodes can be represented by a vector

162: $\mathbf{S}$, with $S_i=1$ when the node $i$ is \textit{on} and $S_i=0$ if

163: the node is \textit{off}.

164: The full phase space containing $2^N$ states is denoted by $\Lambda$.

165: 

166: \subsection{Boolean dynamics} A simple dynamical rule that

167: characterizes the temporal evolution of the state variable can be

168: defined following \cite{tang_1}, which is closely related to update

169: rules applied in perceptron models. 

170: If the network is in the state $\mathbf{S}(t)$ at time $t$, the 

171: state at the next time-step $\mathbf{S}(t+1)$ is given by:

172: 

173: \bq

174: S_i(t+1)=\left\{\begin{array}{ccc}

175: 1 &\text{if}&\sum_j A_{ij}S_j(t)>0\\

176: S_i(t)&\text{if}&\sum_j A_{ij}S_j(t)=0\\

177: 0 &\text{if}&\sum_j A_{ij}S_j(t)<0\\

178: \end{array}\right.

179: \eq

180: 

181: For a given network, we apply this rule to every possible initial

182: condition in $\Lambda$. This defines orbits (trajectories) that must

183: end in a limit cycle (periodic attractor) since we are dealing with a

184: dynamical system on a finite space. A fixed point is a cycle of length one.

185: 

186: Accordingly, $\Lambda$ can be decomposed into a disjoint union of $K$ basins

187: of attraction $B_k$ of size $d_k$: $\Lambda=\bigcup_{k=1}^KB_k$.

188: 

189: In a biological network, the attractors correspond to functional endpoints, and it is important that the states in the attractors are 

190: consistent with observed data.

191: For example, by far the largest endpoint in the cell-cycle network of

192: Li et al.~(see appendix) corresponds to the stationary G1 phase in the cycle.

193: Other systems are more switch-like, for instance in signal transduction,

194: where a cell might change its state from growth to differentiation according 

195: to an external trigger.

196: To characterize these attractors, we introduce the following definitions:

197: 

198: \begin{itemize}

199: \item

200: We compute the \textit{number of attractors} $K$: an

201: attractor is a limit cycle or a fixed point. An attractor $A$ has a basin of

202: attraction $B$ which is the set of all initial conditions whose orbit

203: converges to $A$.

204: 

205: \hspace{-1cm}

206: \item

207: The \textit{basin entropy} $H$ is defined as follows:

208: let $p_k=2^{-N}d_k$ be the probability that an initial state belongs

209: to basin $B_k$. Then, the entropy reads 

210: \be

211: H:=-\sum_{k=1}^K p_k\,\log\left(p_k\right)

212: \ee

213: $H$ is maximum ($H=\log(K)$) if each state is its own basin of size

214: one, and minimum ($H=0$) when there is one single basin.

215: $H$ is a natural measure for characterizing basin structures \cite{book}.

216: Because it takes into account the relative basin sizes, it is quite

217: insensitive to appearance of small and biologically irrelevant

218: basins. 

219: 

220: \item The \textit{perturbation size} $\Delta$

221: measures the distance between attractors of a perturbed and a reference

222: network: from every initial conditions, 

223: the Hamming distance between the fixed points is computed, and the

224: average over all initial conditions is taken. More precisely, if

225: $\mathbf{FP}_G(\mathbf{S})$ 

226: is the fixed point of the trajectory starting at $\mathbf{S}$ and

227: generated by the network G, then 

228: \begin{equation}

229: \Delta_{G,G'}:=\frac{1}{2^N}\sum_{\mathbf{S}}

230: \mbox{HAM}(\mathbf{FP}_G(\mathbf{S}),\mathbf{FP}_{G'}(\mathbf{S}))

231: \label{Delta_def}

232: \end{equation}

233: where $\mbox{HAM}(\cdot,\cdot)$ 

234: is the Hamming distance between two boolean states, namely

235: \begin{equation}

236: \mbox{HAM}(\mathbf{S},\mathbf{T}):=\frac{1}{N}\sum_i |S_i-T_i|

237: \end{equation}

238: 

239: The value $\Delta$ has the following interpretation: it is the average

240: probability (taken over all nodes)  that, for a random initial

241: condition, the final state of a node differs.

242: In this study, the reference network $G$ will be the cell-cycle network of

243: Li et al., which has one very a large basin of attraction and several

244: smaller ones. If some trajectories in the perturbed networks $G'$ end in 

245: a limit cycle, $\Delta$ is defined as the average of the Hamming distance

246: along the cycle.

247: 

248: %we use the following natural generalization:

249: %\begin{eqnarray*}

250: %&&\Delta_{G,G'}:=\\

251: %&&\frac{1}{2^N}\sum_{\mathbf{S}}\sum_{j}

252: %\frac{\mbox{HAM}(\mathbf{FP}_G(\mathbf{S}),\mathbf{LC}_{G'}^{(j)}(\mathbf{S}))}

253: %{|\mathbf{LC}_{G'}(\mathbf{S})|}

254: %\label{Delta_def}

255: %\end{eqnarray*}

256: %where $\mathbf{LC}^{(j)}_G(\mathbf{S})$ is the $j$-th state in the limit cycle,

257: %and $|\mathbf{LC}_{G'}(\mathbf{S})|$ is the length of the cycle.

258: 

259: %Therefore, we slightly modify the definition of $\Delta$

260: %by considering only the main fixed point in the reference network, setting 

261: %$\mathbf{LC}^{(j)}_{G'}(\mathbf{S})=\mbox{const}$ in (\ref{Delta_def}).

262: \end{itemize}

263: 

264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

265: 

266: \subsection{Network models and perturbations}

267: 

268: Our goal is to assess how network dynamics is affected by several types of

269: perturbations.  We consider two classes: one which randomizes the adjacency matrix

270: while keeping a number of topological characteristics from the original network invariant.

271: The second class mimics biological perturbations, as would occur for example through

272: mutations in the interaction partners that constitute the network links.The two classes are defined as follows:

273: 

274: \begin{itemize}

275: \item \textit{Shuffle} (class I): all activating and inhibiting arrows are cut in half and re-wired randomly. This ensure that the connectivity at each node is conserved.

276: As compared to the Li et al.~\cite{tang_1} study, we generate random networks that are more constrained, since the

277: connectivity at each node is forced to remain unchanged after randomization.

278: Such perturbations are applied in the studies of network motifs

279: \cite{alon_8,alon_9}.

280: 

281: \item \textit{Remove} (class II): the arrows are simply suppressed. We extend this class of perturbations beyond single link removal.

282: 

283: \end{itemize}

284: 

285: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

286: 

287: \section{Results}

288: 

289: We study the yeast cell-cycle network of Li et al.~\cite{tang_1}

290: (the \textit{Yeast cell-cycle network} or \textit{YCC}),

291: in which a boolean model reproducing the different phases of

292: the cycle is constructed (see appendix).

293: This model has a main fixed point attracting $86\%$ of the intial

294: conditions.

295: Biologically this state corresponds to the G1 stationary phase of the

296: cell-cycle, as reflected by the activities of the respective nodes.

297: Using computer simulations,

298: the authors further showed that the cell-cycle dynamics had certain

299: robustness properties when challenged with perturbations.

300: In particular, it was shown that in a majority of cases,

301: removal of one link or addition of a link at random did not change much the size of the largest basin of attraction.

302: Finally, the studied network had unusual trajectory channeling

303: properties, when compared to random networks with equal number

304: of nodes and links.

305: Here we extend the characterization of this model by introducing a

306: combination of measures to characterize the structure of basins of

307: attraction as they are modified by structural perturbations. In particular we

308: investigate the consequences of combined mutations and show that they can lead

309: to cancellation effect.

310: 

311: \subsection{Study of shuffled networks (Class I perturbations)}

312: 

313: This type of perturbation allows to study the dynamical characteristics

314: of a biological network in comparison with random networks belonging to a

315: topological class.

316: Figure 1 shows the Number of attractors ($K$) and the Entropy ($H$) of the YCC

317: and randomly shuffled (Class I) versions thereof. 

318: 

319: \begin{figure}[h]

320: %\hspace{-1cm}

321: \includegraphics[width=9cm,angle=0]{NB_Ent}

322: \label{fig_shuff}

323: \caption{Entropy vs. number of attractors after class I perturbation (shuffled arrows). The range of possible $H$ values is indicated by the dashed gray lines. The open red circle represents the reference network, the other points 

324: show the perturbed networks.}

325: \end{figure}

326: 

327: The location of the reference network in the $H-K$

328: plane respective to the scatter of the perturbed networks allows us to asses how

329: typical a network behaves with respect to a class.

330: Accordingly, the YCC is atypical, as seen by its marginal location in the lower left

331: corner. Indeed, this network has lower entropy and fewer basins than most networks, consistent

332: with \cite{tang_1}.

333: 

334: \subsection{Study of mutated networks (Class II perturbations)}

335: 

336: The previous discussion shows how entropy characterizes the system

337: of attractors. However, $H$ contains only information about the relative

338: weights of the attractors, irrespective of their biological relevance.

339: For example a perturbation can decrease the entropy while shifting the

340: fixed point away from from that in the unperturbed, biologically

341: relevant state.

342: For this reason we introduced a second quantity, $\Delta$ (Equation

343: \ref{Delta_def}), a probabilistic measure of the change in the

344: fixed point after perturbation. Therefore, $\Delta$ reflects the

345: change in the biological relevance of the basin structure.

346: 

347: \begin{figure}[h]

348: \centering

349: \includegraphics[width=4cm,angle=0]{NB_Ent_rmv_nb}

350: \includegraphics[width=3.8cm,angle=0]{D_H_rmv_nb}

351: \includegraphics[width=9cm,angle=0]{Ent_D_rmv_nb}

352: \label{fig_rmv}

353: \caption{Entropy, number of attractors and $\Delta$ after class II perturbation

354: (removed arrows). Colors represent different number of removed arrows:

355: black for one removed arrows, red 

356: for 2, green for 3, turquoise for 4 and yellow for more than 4.

357: A: same figure as for the class I perturbation,

358: the range of possible $H$ values is indicated by the dashed gray lines,

359: the open blue circles represent the reference network.

360: B: Distribution of $\Delta$.

361: C: $\Delta$ vs. $H$ plot, the dashed gray line represents the

362: entropy of the reference network.}

363: \end{figure}

364: 

365: We first repeat Figure 1 for class II perturbations which shows that networks with few perturbations

366: cluster around the wild-type model (Figure 2A), while the sread for networks with four perturbations resembles

367: the shuffled models (Figure 1). Turning to the measure of $\Delta$, we find that

368: $\Delta$-distribution

369: (Figure 2B) is bimodal, showing two distinct populations of perturbations:

370: ($\Delta\lesssim 0.2$ and $\Delta\gtrsim 0.2$).

371: In the second case, the perturbed

372: model does not reproduce the biologically correct cell-cycle progression.

373: But if $\Delta$ is small, then the system of attractors of the perturbed network

374: is still consistent with the biology and entropy allows to

375: discriminate between networks with a larger or smaller main basin of attraction.

376: For this reason, the entropy and $\Delta$ are complementary for describing

377: the dynamical landscape (Figure 2C).

378: The two different modes in the $\Delta$-histogram are clearly reflected on

379: this 2D representation. Noticeably, the $\Delta$ values span a broad range

380: for any number of removed arrows, on the other hand higher entropies

381: are more frequent for larger number ($>2$) of removed arrows.

382: Qualitatively, the spread of points in the $H-\Delta$ plane conveys

383: a measure of \textit{robustness}. Accordingly, the $\Delta$ measure appears

384: more fragile than the entropy property, especially when few arrows are removed.

385: 

386: We now interpret the different locations in the  $H-\Delta$ plane:\\

387: %\begin{enumerate}

388: %\item

389: 1. If $\Delta$ is large ($\Delta\gtrsim 0.2$), the model does have attractor states which coincide with the gene activities

390: of the different cell-cycles phases. Such perturbations are specially 

391: interesting if the number of removed arrows is small (dark colors).

392: Such links are then essential for the model, as their removal disrupts the cell-cycle very efficiently.\\

393: %\item

394: 2. If $\Delta$ is small and the entropy increases,

395: the probability that the dynamics ends in the reference attractor

396: decreases demonstrating that the removed arrows contributed

397: to the channeling properties of the system.\\

398: %\item

399: 3. If $\Delta$ is small and the entropy decreases,

400: the main attractor of the perturbed network has a stronger attraction property.

401: Some of these networks could be considered as alternative cell-cycle models.\\

402: %\end{enumerate}

403: 

404: We illustrate these three regimes by examples:

405: 

406: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

407: \subsection{Examples of mutations}

408: 

409: In the first example (Table II), the dynamics has a large main basin of attraction like in the unperturbed model (Table I). However, 

410: the fixed point is significantly different from wild-type as

411: the system is blocked in the a state of the M-phase and

412: cannot finish properly the cell-cycle

413: (see Appendix for the recapitulation of the wild-type mode from \cite{tang_1}).

414: 

415: \begin{table}[h]

416: {\tiny

417: \begin{tabular}{cccccccccccl}

418: Cln3 & MBF & SBF & Cln1,2 & Cdh1 & Swi5 & Cdc20,14 & Clb5,6 & Sic1 & Clb1,2 & Mcm1 & $\%$ \\

419: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.8613 \\

420: 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0737 \\

421: 0 & 1 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0532 \\

422: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0043 \\

423: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0034 \\

424: 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0034 \\

425: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0004 \\

426: \end{tabular}}

427: \caption{

428: Basins of attraction with their respective probabilities in ($\%$) for the original YCC network. The largest basin ends at the G1 stationary state. Entropy $H= 0.543$, Number of attractors $K=7$.}

429: \end{table}

430: 

431: \begin{table}

432: {\tiny

433: \begin{tabular}{cccccccccccl}

434: Cln3 & MBF & SBF & Cln1,2 & Cdh1 & Swi5 & Cdc20,14 & Clb5,6 & Sic1 & Clb1,2 & Mcm1 & $\%$ \\

435: 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 1 & 1 & 1 & 0.880 \\

436: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.054 \\

437: 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.027 \\

438: 0 & 1 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.015 \\

439: 0 & 0 & 0 & 0 & 1 & 1 & 1 & 0 & 1 & 1 & 1 & 0.010 \\

440: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.004 \\

441: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.003 \\

442: 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.003 \\

443: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0.000 \\

444: \end{tabular}}

445: \caption{

446: Basins of attraction with their respective probabilities, when (Cdc20,Cdc14) $\rightarrow$ Clb1,2 and Sic1 $\rightarrow$ Clb1,2 are removed. Entropy = 0.549, Number of attractors = 9, $\Delta$ = 0.41.}

447: \end{table}

448: 

449: In the second example (Table III), the dynamics has the same main fixed point as

450: the wild-type, but with a smaller basin of attraction, while the

451: second biggest has grown. Therefore the removed connection SBF

452: $\rightarrow$ Cln1,2 contributes to the ability of the main fixed point to funnel trajectories.

453: 

454: 

455: \begin{table}[h]

456: {\tiny

457: \begin{tabular}{cccccccccccl}

458: Cln3 & MBF & SBF & Cln1,2 & Cdh1 & Swi5 & Cdc20,14 & Clb5,6 & Sic1 & Clb1,2 & Mcm1 & $\%$\\

459: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.6669 \\

460: 0 & 1 & 1 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.1762 \\

461: 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0654 \\

462: 0 & 1 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0532 \\

463: 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0180 \\

464: 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0043 \\

465: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0043 \\

466: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0034 \\

467: 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0034 \\

468: 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0034 \\

469: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0004 \\

470: 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0004 \\

471: \end{tabular}}

472: \caption{

473: Basins of attraction with their respective probabilities, when SBF $\rightarrow$ Cln1,2 is removed. Entropy $H=1.096$, Number of attractors $K=12$, $\Delta$ = 0.05.}

474: \end{table}

475: 

476: The third example (Table IV) is a model with four removed arrows which has the same

477: main fixed point with a slightly higher probability. Also, the second 

478: largest fixed point is same as in the wild-type model. This indicates that the effect

479: of some mutations can be canceled by further mutations.

480: While such cases exist, we found that networks with several removed links that

481: preserving the unperturbed cell-cycle behavior are rare.

482: 

483: \begin{table}[h]

484: {\tiny \vspace{0.5cm}

485: \begin{tabular}{cccccccccccl}

486: Cln3 & MBF & SBF & Cln1,2 & Cdh1 & Swi5 & Cdc20,14 & Clb5,6 & Sic1 & Clb1,2 & Mcm1 & $\%$\\

487: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.8793 \\

488: 0 & 0 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0507 \\

489: 0 & 1 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0356 \\

490: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0268 \\

491: 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0.0034 \\

492: 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0034 \\

493: 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0.0004 \\

494: \end{tabular}}

495: \caption{

496: Basins of attraction with their respective probabilities, when (Cdc20,Cdc14) $\rightarrow$ Clb1,2, Clb1,2 $\rightarrow$ Mcm1, Clb1,2 $\rightarrow$ Cdh1 and Clb1,2 $\rightarrow$ Swi5 are removed.

497: Entropy $H= 0.523$, Number of attractors $K=7$, $\Delta$ = 0.025.}

498: \end{table}

499: 

500: \vspace{-0.5cm}

501: \section{Conclusion}

502: 

503: We have proposed a systematic approach for studying the dynamical

504: attractor landscape of biological networks, and their response to structural

505: perturbations. In particular, we introduced a low dimensional representation

506: of the system of attractors, the entropy, and a probabilistic measure

507: in the perturbation size $\Delta$.

508: This enabled us to study the global characteristics of network

509: perturbation in a compact and visually effective form.

510: In a biological context, this can provide hints to elucidate the dynamical role of

511: specific network links. Alternatively, the function of new and yet unobserved links can

512: be predicted as in \cite{stoll}, and imperfect starting models can be improved.

513: 

514: We applied this method to a model of the yeast cell-cycle by Li et al.

515: Using the measures introduced, we have generalized the dynamical characterization of

516: the model using a broad range of perturbations.

517: This has enabled us to emphasize the breadth of

518: dynamical behavior (Figure 2) induced by only few mutated links.

519: Interestingly, 

520: we observed (Figure 2C) that the structure of the system of attractors ($H$) behaves quite robustly

521: compared to the modification in the final states ($\Delta$), especially when the number of removed links

522: is small ($<3$).

523: We illustrated through examples the consequences of removing

524: individual or groups of links. Interestingly it was possible to remove up to four   links while

525: not affecting the basin structure significantly. 

526: Tracking the dynamical changes in the activity levels of proteins in a network is a very high-dimensional problem. 

527: It therefore important to be have few informative variables which allow one to efficiently assess

528: a large number of perturbed models at once.

529: We believe that basin entropy and distance to a reference attractor are well suited for this purpose.

530: 

531: \subsection*{Acknowledgments}

532: We thank the organizers of the CompBioNets '04 conference (Recife, Brazil) at which an initial version of this work was presented. The simulations were performed on an Itanium2 cluster from HP/Intel at the Vital-IT facilities.

533: FN ad GS acknowledge funding from the NCCR Molecular Oncology program and NIH administrative supplement to parent

534: grant GM54339.

535: 

536: \appendix

537: \section{the Yeast cell-cycle network of Li et al.}

538: The following two tables are recapitulated from \cite{tang_1}.

539: 

540: \begin{table}[h]

541: {\tiny

542: \begin{tabular}{cccccccccccc}

543: 1$\stackrel{+}{\rightarrow}$2&

544: 1$\stackrel{+}{\rightarrow}$3&

545: 2$\stackrel{+}{\rightarrow}$8&

546: 3$\stackrel{+}{\rightarrow}$4&

547: 6$\stackrel{+}{\rightarrow}$9&

548: 7$\stackrel{+}{\rightarrow}$5&

549: 7$\stackrel{+}{\rightarrow}$6&

550: 7$\stackrel{+}{\rightarrow}$9&

551: 8$\stackrel{+}{\rightarrow}$10&

552: 8$\stackrel{+}{\rightarrow}$11&

553: 10$\stackrel{+}{\rightarrow}$7&

554: 10$\stackrel{+}{\rightarrow}$11\\

555: 

556: 11$\stackrel{+}{\rightarrow}$6&

557: 11$\stackrel{+}{\rightarrow}$7&

558: 11$\stackrel{+}{\rightarrow}$10&

559: 4$\stackrel{-}{\rightarrow}$9&

560: 4$\stackrel{-}{\rightarrow}$5&

561: 5$\stackrel{-}{\rightarrow}$10&

562: 7$\stackrel{-}{\rightarrow}$8&

563: 7$\stackrel{-}{\rightarrow}$10&

564: 8$\stackrel{-}{\rightarrow}$5&

565: 8$\stackrel{-}{\rightarrow}$9&

566: 9$\stackrel{-}{\rightarrow}$8&

567: 9$\stackrel{-}{\rightarrow}$10\\

568: 

569: 10$\stackrel{-}{\rightarrow}$2&

570: 10$\stackrel{-}{\rightarrow}$3&

571: 10$\stackrel{-}{\rightarrow}$5&

572: 10$\stackrel{-}{\rightarrow}$6&

573: 10$\stackrel{-}{\rightarrow}$9&

574: 1$\stackrel{-}{\rightarrow}$1&

575: 4$\stackrel{-}{\rightarrow}$4&

576: 6$\stackrel{-}{\rightarrow}$6&

577: 7$\stackrel{-}{\rightarrow}$7&

578: 11$\stackrel{-}{\rightarrow}$11\\

579: \end{tabular}}

580: \caption{Adjacency matrix of the Yeast cell-cycle network. The numbers refer to the ordering of the nodes as used in Tables I-IV,VI. $+$ (respectively $-$)

581: represent activating (respectively repressing) links.  }

582: \end{table}

583: 

584: %\subsection{Cell-cycle states of the model of Li et al~\cite{tang_1}.}

585: 

586: \begin{table}[h]

587: {\tiny

588: \begin{tabular}{ccccccccccccc}

589:  t & Cln3 & MBF & SBF & Cln1,2 & Cdh1 & Swi5 & C20,14 & Clb5,6 & Sic1 & Clb1,2 & Mcm1 & Phase \\ \\

590: 1  & 1  & 0  & 0  & 0  & 1  & 0  & 0  & 0  & 1  & 0  & 0  & START \\

591: 2  & 0  & 1  & 1  & 0  & 1  & 0  & 0  & 0  & 1  & 0  & 0  & G1 \\

592: 3  & 0  & 1  & 1  & 1  & 1  & 0  & 0  & 0  & 1  & 0  & 0  & G1 \\

593: 4  & 0  & 1  & 1  & 1  & 0  & 0  & 0  & 0  & 0  & 0  & 0  & G1 \\

594: 5  & 0  & 1  & 1  & 1  & 0  & 0  & 0  & 1  & 0  & 0  & 0  & S \\

595: 6  & 0  & 1  & 1  & 1  & 0  & 0  & 0  & 1  & 0  & 1  & 1  & G2 \\

596: 7  & 0  & 0  & 0  & 1  & 0  & 0  & 1  & 1  & 0  & 1  & 1  & M \\

597: 8  & 0  & 0  & 0  & 0  & 0  & 1  & 1  & 0  & 0  & 1  & 1  & M \\

598: 9  & 0  & 0  & 0  & 0  & 0  & 1  & 1  & 0  & 1  & 1  & 1  & M \\

599: 10  & 0  & 0  & 0  & 0  & 0  & 1  & 1  & 0  & 1  & 0  & 1  & M \\

600: 11  & 0  & 0  & 0  & 0  & 1  & 1  & 1  & 0  & 1  & 0  & 0  & M \\

601: 12  & 0  & 0  & 0  & 0  & 1  & 1  & 0  & 0  & 1  & 0  & 0  & G1 \\

602: 13 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 0 & G1* \\

603: \end{tabular}}

604: \caption{This table represents the discrete time evolution of the boolean states of the YCC network as it traverses the different cell-cycle phases. Cdc20.14 has been abbreviated C20,14; G1* indicates the stationary G1 phase. }

605: \end{table}

606: 

607: \vspace{7cm}

608: 

609: \begin{thebibliography}{}

610: %systems's biology

611: \bibitem{leibler_3}

612: Hartwell, L. H., Hopfield, J. J., Leibler, S., and Murray, A. W. (1999). From molecular to modular cell biology, Nature 402, C47-52.

613: \bibitem{arkin_4}

614: Alm, E., and Arkin, A. P. (2003). Biological networks, Curr Opin Struct Biol 13, 193-202.

615: \bibitem{barabasi_3}

616: Oltvai, Z. N., and Barabasi, A. L. (2002). Systems biology. Life's complexity pyramid, Science 298, 763-4.

617: 

618: %networks

619: \bibitem{alon_8}

620: Alon, U. (2003). Biological networks: the tinkerer as an engineer, Science 301, 1866-7.

621: \bibitem{barabasi_2}

622: Barabasi, A. L., and Oltvai, Z. N. (2004). Network biology: understanding the cell's functional organization, Nat Rev Genet 5, 101-13.

623: 

624: %networks, scaling  etc

625: \bibitem{albert_barabasi_2}

626: Barabasi, A. L., and Albert, R. (1999). Emergence of scaling in random networks, Science 286, 509-12.

627: 

628: %motifs

629: \bibitem{alon_9}

630: Milo, R., Shen-Orr, S., Itzkovitz, S., Kashtan, N., Chklovskii, D., and Alon, U. (2002). Network motifs: simple building blocks of complex networks, Science 298, 824-7.

631: 

632: %modules

633: \bibitem{barkai_9}

634: Ihmels, J., Friedlander, G., Bergmann, S., Sarig, O., Ziv, Y., and Barkai, N. (2002). Revealing modular organization in the yeast transcriptional network, Nat Genet 31, 370-7.

635: %noise

636: \bibitem{arkin2_4}

637: Arkin, A., Ross, J., and McAdams, H. H. (1998). Stochastic kinetic analysis of developmental pathway bifurcation in phage lambda-infected Escherichia coli cells, Genetics 149, 1633-48.

638: 

639: %modeling

640: \bibitem{tyson_novak_5} Novak, B., Csikasz-Nagy, A., Gyorffy, B., Chen, K., and Tyson, J. J. (1998). Mathematical model of the fission yeast cell cycle with checkpoint controls at the G1/S, G2/M and metaphase/anaphase transitions, Biophys Chem 72, 185-200.

641: \bibitem{cross_3}

642: Cross, F. R., Archambault, V., Miller, M., and Klovstad, M. (2002). Testing a mathematical model of the yeast cell cycle, Mol Biol Cell 13, 52-70.

643: \bibitem{leibler_barkai_1}

644: Barkai, N., and Leibler, S. (2000). Circadian clocks limited by noise, Nature 403, 267-8.

645: \bibitem{leibler_2}

646: Vilar, J. M., Kueh, H. Y., Barkai, N., and Leibler, S. (2002). Mechanisms of noise-resistance in genetic oscillators, Proc Natl Acad Sci U S A 99, 5988-92. Epub 2002 Apr 23.

647: \bibitem{goldbeter_1}

648: Leloup, J. C., and Goldbeter, A. (2003). Toward a detailed computational model for the mammalian circadian clock, Proc Natl Acad Sci U S A 100, 7051-6. Epub 2003 May 29.

649: \bibitem{von_dassow_1}

650: von Dassow, G., Meir, E., Munro, E. M., and Odell, G. M. (2000). The segment polarity network is a robust developmental module, Nature 406, 188-92.

651: \bibitem{yuhai_tu_1}

652: Mello, B. A., and Tu, Y. (2003). Quantitative modeling of sensitivity in bacterial chemotaxis: the role of coupling among different chemoreceptor species, Proc Natl Acad Sci U S A 100, 8223-8. Epub 2003 Jun 25.

653: \bibitem{yuhai_tu_2}

654: Mello, B. A., and Tu, Y. (2003). Perfect and near-perfect adaptation in a model of bacterial chemotaxis, Biophys J 84, 2943-56.

655: \bibitem{levchenko_1}

656: Hoffmann, A., Levchenko, A., Scott, M. L., and Baltimore, D. (2002). The IkappaB-NF-kappaB signaling module: temporal control and selective gene activation, Science 298, 1241-5.

657: \bibitem{brown_sethna_1}

658: Brown, K. S., and Sethna, J. P. (2003). Statistical mechanical approaches to models with many poorly known parameters, Phys Rev E Stat Nonlin Soft Matter Phys 68, 021904. Epub 2003 Aug 12.

659: 

660: %boolean

661: \bibitem{kadanoff}

662: Aldana, A., Coppersmith, S., Kadanoff, L.P. (2002), Boolean dynamics with random couplings (http://arxiv.org/abs/nlin/0204062).

663: \bibitem{kauffman_1}

664: Kauffman, S., Peterson, C., Samuelsson, B., and Troein, C. (2003). Random Boolean network models and the yeast transcriptional network, Proc Natl Acad Sci U S A 100, 14796-9. Epub 2003 Dec 1.

665: \bibitem{tang_1}

666: Li, F., Long, T., Lu, Y., Ouyang, Q., and Tang, C. (2004). The yeast cell-cycle network is robustly designed, Proc Natl Acad Sci U S A 101, 4781-6. Epub 2004 Mar 22.

667: \bibitem{albert_3}

668: Albert, R., and Othmer, H. G. (2003). The topology of the regulatory interactions predicts the expression pattern of the segment polarity genes in Drosophila melanogaster, J Theor Biol 223, 1-18.

669: \bibitem{albert_4}

670: Chaves, M., Albert, R., and Sontag, E. D. (2005). Robustness and fragility of Boolean models for genetic regulatory networks, J Theor Biol 235, 431-449.

671: 

672: %eckmann

673: \bibitem{eckmann}

674: Vazquez, A., Dobrin, R., Sergi, D., Eckmann, J.-P., Oltvai, Z. N., and

675: Barabasi, A.-L. (2004). The topological relationship between the

676: large-scale attributes and local interaction patterns of complex

677: networks, Proc Natl Acad Sci U S A 101, 17940-5.

678: 

679: %stoll

680: \bibitem{stoll}

681: Stoll G., Rougemont J. and Naef F. (2006). Few crucial links assure checkpoint efficiency in the yeast cell-cycle network,

682: Bioinformatics 22(20):2539-46.

683:  

684: %books

685: \bibitem{book}

686: Cover, T. M. and Thomas, J.A. (1991). Elements of information theory, Wiley.

687: 

688: \end{thebibliography}

689: \end{document}

690: