1: %\documentstyle[preprint,aps,psfig,epsf]{revtex}

2: \documentclass[twocolumn,aps,prb]{revtex4}

3: \usepackage{epsf,graphicx}

4: 

5: \begin{document}

6: \draft

7: 

8: \title{Finite size effects on thermal denaturation of globular proteins}

9: \author{Mai Suan Li$^1$, D. K. Klimov$^2$ and D. Thirumalai$^2$}

10: 

11: \address{$^1$Institute of Physics, Polish Academy of Sciences,

12: Al. Lotnikow 32/46, 02-668 Warsaw, Poland\\

13: $^2$Department of Chemistry and Biochemistry and Institute for Physical

14: Science and Technology, University of Maryland, College Park, MD 20742 }

15: 

16: \begin{abstract}

17: 

18: Finite size effects on the cooperative thermal denaturation of proteins

19: are considered. A dimensionless measure of cooperativity, $\Omega_c$, 

20: scales as $N^\zeta$, where $N$ is the number of amino acids. Surprisingly,

21: we find that $\zeta$ is universal with $\zeta = 1 + \gamma$, where the

22: exponent $\gamma$ characterizes the divergence of the susceptibility for a

23: self-avoiding walk. Our lattice model simulations and 

24: experimental data are consistent with the theory. Our finding rationalizes

25: the marginal stability of proteins and substantiates the earlier predictions

26: that the efficient folding of two-state proteins requires $T_F\approx

27: T_\theta$, where $T_\theta$ and $T_F$ are the collapse and folding

28: transition temperatures, respectively.

29: 

30: \end{abstract}

31: 

32: \maketitle

33: 

34: %\pacs{PACS numbers: 87.10.+e,87.15.Cc,36.20.-r}

35: 

36: %]

37: 

38: 

39: Single domain globular proteins, which are finite sized systems, undergo

40: remarkably cooperative transitions from an ensemble of unfolded states to

41: well ordered folded (or native) states as the temperature is lowered (Fig.

42: 1(a)).  In many cases, the transition to the native state takes place in

43: an apparent two-state manner, i.e., the only detectable species are the

44: native (more precisely, the conformations belonging to the

45: native basin of attraction ({\bf NBA})) or unfolded ({\bf U}) states

46: \cite{Poland}.  Although the microscopic origin of cooperativity is

47: not fully understood \cite{Bryngelson95Proteins}, the transition to the

48: {\bf NBA} at the folding transition temperature, $T_F$, 

49: is a consequence of the effective interresidue attraction that

50: compensates for the entropy loss.  From this

51: perspective the {\bf NBA} $\leftrightarrow$ {\bf U} transition can be

52: viewed as a phase transition in a finite-sized system. 

53: Furthermore, the transition to the {\bf NBA} at $T_F$

54: has the characteristics of a first order phase transition

55: \cite{Poland,Bryngelson95Proteins}.  Many experiments have shown that

56: folded states of globular proteins are only marginally stable below $T_F$.  

57: The free energies of stability of the {\bf NBA}, relative to the {\bf U}

58: states, vary within the range of $(5 - 20)k_BT$ at neutral pH [1b].

59: Because proteins are polymers we

60: expect that they would also undergo a collapse transition to a compact

61: phase at the temperature $T_{\theta}$ suitably modified for 

62: finite size systems, when water becomes a poor solvent

63: for the polypeptide. We have previously shown that for protein sequences

64: that fold in an apparent two-state manner $T_F \approx T_{\theta}$, which

65: naturally explains the marginal stability of proteins

66: \cite{Camacho93PNAS}.

67: 

68: The quest to understand, at the molecular level, the cooperative {\bf U}

69: $\leftrightarrow$ {\bf NBA} transition has lead to a number of

70: computational studies [2b,4,5].  

71: %\cite{Skolnick96Proteins,Scheraga98JMB,KlimThirum98FD}.  

72: Although considerable effort has been directed to describe the molecular

73: basis of cooperativity, somewhat surprisingly, examination of the finite size

74: effects in the self-assembly of proteins has received little attention

75: \cite{KlimThirum02JCC}. In

76: contrast, scaling theories for finite sized systems undergoing regular

77: first and second order phase transitions have been fully developed

78: \cite{Fisher}.   The purpose of this paper

79: is to study the effect of $N$, the number of amino acid residues in a

80: protein, on the extent of cooperativity in the {\bf U} $\leftrightarrow$

81: {\bf NBA} transition.

82: 

83: Thermal denaturation data of wild-type (WT)  

84: proteins and lattice models (LMs) 

85: of polypeptide chains are used to examine the 

86: dependence of the cooperativity on 

87: $N$. We show that a dimensionless measure of cooperativity

88: \cite{KlimThirum98FD}

89: \begin{equation}

90: \Omega _c \; = \; \frac{T_F^2}{\Delta T}

91: |\frac{df_N}{dT}| _{T=T_F}

92: \label{coop_eq}

93: \end{equation}

94: grows as

95: \begin{equation}

96: \Omega _c \; \sim \; N^{\zeta}, 

97: \label{coop_scaling_eq}

98: \end{equation}

99: where $f_N$ is a measure of occupation of {\bf NBA}, $\Delta T$ is the full

100: width at half-maximum of $df_N /dT$, and $T_F$ is the folding

101: transition temperature identified with the maximum in  $df_N /dT$. 

102: We find that

103: \begin{equation}

104: \zeta \; = \; 1 + \gamma

105: \label{zeta_exp_eq}

106: \end{equation} 

107: where $\gamma$ is the exponent that characterizes the

108: divergence of susceptibility at the critical point for a $n$-component

109: ferromagnet with $n=0$, i.e., for a self-avoiding walk. As a byproduct of

110: this study we also show that $\frac{\Delta T}{T_F} \sim \frac{1}{N}$. The

111: parameter $\Omega_c$ is a convolution of the sharpness of the transition

112: ($T_F/\Delta T$) and the extent to which structure, as measured by $f_N$,

113: changes around $T_F$. For infinite systems undergoing sharp transitions,

114: $\Omega _c \rightarrow \infty$, whereas $\Omega _c$ is small for broad or

115: highly rounded phase transitions \cite{KlimThirum98FD}. The relationship

116: given in Eq. (\ref{zeta_exp_eq}), which can only be valid near $T_\theta$,

117: establishes the proposal that $T_F\approx T_\theta$ for two-state

118: folders \cite{Camacho93PNAS,Thirum95}.

119:  

120: 

121: To establish the results given above we  used thermal and chemical

122: denaturation data together with  the LMs 

123: of a polypeptide chain to compute the growth of $\Omega _c$ with $N$. 

124: In the LM each amino acid is represented as a single

125: bead confined to the vertices of a cubic lattice [2c]. The energy of a

126: conformation specified by the positions, $\{\vec{r}_i\} (i = 1, 2, \ldots ,N)$,

127: is $E\{\vec{r}_i\}=\sum_{i<j} \, \epsilon_{ij} \delta_{r_{ij},a}$, 

128: where $a$ is the lattice spacing, $r_{ij} = |\vec{r}_i - \vec{r}_j|$,

129: $\delta_{x,a}$ is the Kronecker delta function. The contact energies

130: $\epsilon _{ij} = -1$, if the interaction between beads $i$ and $j$ in a

131: given conformation is also present in the native state (i.e., the lowest

132: energy conformation for a given sequence), and is zero,

133: otherwise. Even though simple LMs do not quantitatively 

134: capture the cooperativity of folding transitions in proteins 

135: \cite{Levitt97ProtSci}, they are 

136: useful for obtaining global folding properties.  The precise

137: choice of $\epsilon _{ij}$ should not affect the predicted universal

138: scaling of $\Omega _c$ with $N$.  Our purpose in undertaking LM 

139: Monte Carlo (MC) 

140: simulations is to show that Eqs.  (\ref{coop_eq})-(\ref{zeta_exp_eq})

141: should be valid for any model of proteins that exhibits a cooperative {\bf

142: U} $\leftrightarrow$ {\bf NBA} transition.

143: 

144: To calculate $\Omega _c$ for LMs we employ the temperature

145: dependence of the overlap function \cite{Camacho93PNAS}

146: \begin{equation}

147: \chi \; = \; 1 - \frac{1}{N^2-3N+2} \sum_{i<j+1}^N \,\;\delta_{r_{ij},r_{ij}^0}

148: \label{chi_eq}

149: \end{equation}

150: where $r_{ij}^0$ is the distance between beads $i$ and $j$ in the native

151: conformation. The overlap function $\chi$ 

152: is an order parameter that distinguishes the {\bf NBA} and {\bf U} states.  

153: The folding transition temperature $T_F$ can be estimated from the

154: location of the maximum in $d<\chi>/dT$, where $<...>$ indicates a thermal

155: average.  For LMs $<\chi> \approx 1- f_N$ \cite{LMCalculations}. Therefore, Eq.

156: (\ref{coop_eq}) may be evaluated using

157: \begin{equation}

158: \Omega _c \; = \;  \; \frac{T_F^2}{\Delta T}

159: \biggl(\frac{d<\chi>}{dT}\biggr)_{T=T_F}. 

160: \label{coop_lattice_eq}

161: \end{equation}

162: 

163: 

164: 

165: {\em Analysis of experimental and simulation

166: data}: To establish the results given above we

167: first analyzed thermal denaturation data for WT proteins.  

168: As an example we show in Fig. \ref{fN_fig} the plot of $f_N(T)$ and

169: $df_N(T)/dT$ for villin ($N=35$) and ADA 2h ($N=80$)

170: \cite{ProteinDB}. In accord with Eq. (\ref{coop_scaling_eq}) we find that the

171: thermal denaturation of ADA 2h is more cooperative than that of

172: villin headpiece.

173: 

174: %FIGURE 1

175: \begin{figure}

176: \epsfxsize=4in

177: \centerline{\epsffile{Fig1.eps}}

178: \caption{ (a)

179: Temperature dependence of the fraction of occupation of the native state,

180: $f_N(T)$, and its derivative $df_N/dT$.

181: The dotted lines are for villin head piece and the solid lines

182: show the data for ADA 2h \cite{ProteinDB}.

183: Temperature is measured in Centigrades. (b)

184: Dependence of 1-$<\chi>$ and $d<\chi >/dT$ on temperature

185: for LMs. We calculate

186: $\Delta T$ using $d<\chi>/dT$. The dotted lines are for the sequence with

187: $N=48$ and the solid lines correspond to $N=80$.

188: }

189: \label{fN_fig}

190: \end{figure}

191: 

192: % ???

193: 

194: From Fig. 2 we find that

195: $\frac{\Delta T}{T_F}$, from thermal denaturation data for 32 WT

196: proteins \cite{ProteinDB},

197: scales as $N^{-\lambda}$ with $\lambda

198: = 1.08 \pm 0.04$.  Given that the data for

199: these proteins are obtained under varying experimental conditions and

200: using different

201: methods for computing the enthalpy and entropy changes at $T_F$, the

202: agreement between the predicted and observed behavior is excellent.

203: For LMs $\Delta T/T_F \sim

204: N^{-\lambda}$ with $\lambda=1.14\pm 0.06$ (Fig. 2). The

205: small deviation

206: of $\lambda$ from unity in LMs  is, in all likelihood, due to

207: the simplicity of the $\alpha$-carbon representation of the polypeptide

208: chain that does not capture the crucial role of side chains. 

209: Inclusion of side chains, which are 

210: tightly packed in native conformations, is expected to reduce 

211: fluctuations. Moreover, for $N\lesssim 40$ most of the beads are on the

212: surface, which also leads to considerable conformational fluctuations.  

213: Therefore, the expected relation $\frac{\Delta T}{T_F}

214: \sim N^{-1}$ holds nearly quantitatively. 

215: 

216: %FIGURE 2

217: \begin{figure}

218: \epsfxsize=3.5in

219: \centerline{\epsffile{Fig2.ps}}

220: \caption{ The sharpness of the folding transition

221: $\Delta T/T_F$ as a function of $N$. Open circles represent the results from

222: LM simulations  with the corresponding fit (dotted line)

223: $\Delta T/T_F \sim N^{-\lambda}$ with $\lambda = 1.14 \pm

224: 0.08$. The linear fit (solid line) to the experimental data for

225: 32 WT proteins (solid circles) \cite{ProteinDB}  gives

226: $\lambda = 1.08 \pm 0.04$. The

227: correlation coefficient for  $\ln \Delta T/T_F$ and $\ln N$ is 0.95.

228: For clarity

229: LM  data are shifted up by 0.4. Inset shows the dependence of the

230: width of folding transition $\ln \Delta C/C_m$

231: for chemical denaturation on $\ln N$. The

232: linear fit to the data points collected for 33 WT  proteins

233: yields $\lambda = 1.22\pm 0.14$ (the correlation factor is 0.59).

234: }

235: \end{figure}

236: 

237: 

238: The dependence of $\Omega_c$ on $N$ for WT  proteins and LMs shows

239: that $\Omega_c \sim N^{\zeta}$ (Fig. (3)).   From the linear fit to

240: the log-log plot of the data  we find $\zeta \approx 2.17\pm 0.09$ for

241: WT proteins and $\approx 2.33\pm 0.08$ for LMs. The 5th order

242: $\epsilon$ expansion for polymers using $n$-component $\phi^4$ theory

243: with $n=0$ gives $\gamma=1.22$ \cite{Kleinert}. Thus, from

244: Eq. (\ref{zeta_exp_eq}) we predict that $\zeta \approx 2.22$.  Thus,

245: the data for  WT proteins and LMs  are consistent with the

246: theoretical prediction (Eq. (\ref{zeta_exp_eq})).  We should emphasize

247: that the robustness of the fit has been checked using different

248: fitting procedures.  The {\em remarkable finding relating the critical

249: exponent $\gamma$ to thermal denaturation of proteins} gives further

250: credence to the proposal that efficient folding is achieved at $T_F

251: \approx T_{\theta}$ \cite{Camacho93PNAS}. 

252: It also suggests that {\bf U} $ \leftrightarrow$

253: {\bf NBA} transition is only weakly first order, thus explaining the

254: marginal stability of globular proteins.

255: 

256: %FIGURE 3

257: \begin{figure}

258: \epsfxsize=3.5in

259: \centerline{\epsffile{Fig3.ps}}

260: \caption{ Plot of ln$\Omega_c$ as a function of ln$N$.

261: Symbols are the same as in Fig. (2).

262: The dotted line is a  fit to the LM data, which gives

263: $\zeta =2.33 \pm 0.08$.

264: The solid line is a fit to the experimental values of

265: $\Omega_c$ \cite{ProteinDB} with the  exponent

266: $\zeta = 2.17 \pm 0.09$. The correlation coefficient for

267: $\ln \Omega_c$  and $\ln N$ is 0.95.

268: The LM data are shifted down by 0.7. The dependence of the folding

269: cooperativity  on $N$ for chemical denaturation is plotted in the

270: inset. The linear fit to experimental data (solid line) results in

271: the exponent $\zeta = 2.45\pm 0.29$ (the correlation factor is 0.59).

272: Both sets of experimental data rule out $\zeta=2$.

273: }

274: \end{figure}

275: 

276: Most folding experiments are performed by titrating with

277: denaturants (urea or guanidine hydrochloride). At denaturant

278: concentrations above the midpoint $C_m$ (at which the populations of the

279: folded and unfolded states are equal) proteins are denaturated. Thus,

280: phase transitions to the {\bf NBA} occur by varying denaturant

281: concentration. In analogy with Eq. (\ref{coop_eq}) we computed, for

282: 33 WT proteins \cite{ProteinDB}, $\Omega _c = \frac{C_m^2}{\Delta C}

283: | \frac{df_N}{dC}| _{C=C_m}$ and $\frac{\Delta C}{C_m}$, where

284: $\Delta C$ is the full width at half-maximum of $df_N/dC$. The plots of

285: ln$\frac{\Delta C}{C_m}$ and ln$\Omega_c$ as a function of ln$N$ yield

286: $\lambda \approx 1.22\pm0.14$ and  

287: $\zeta \approx  2.45\pm0.29$, respectively (see the

288: insets to Figs. (2,3)).  Thus, the

289: scaling of $\Omega_c$ and $\Delta C/C_m$ remains essentially

290: unchanged even {\em  though the

291: chemical and thermal denaturation mechanisms are vastly different}. This

292: result also suggests  that $\zeta$ is universal. However, the dependence

293: of ln$\Omega_c$ on ln$N$ has a correlation coefficient of only about 0.6

294: compared to 0.95 for thermal denaturation. We believe that larger

295: uncertainties result from greater drift in the

296: experimental signals in denaturant-induced unfolding compared to thermal

297: denaturation \cite{Camacho93PNAS}.

298: 

299: 

300: The rationale for Eqs. (2,3) 

301: is based on the following arguments. (1) By analogy with

302: magnetic systems $\Delta \chi$ is similar to susceptibility and should be

303: given by $\Delta \chi = T\partial <\chi>/\partial h$, where $h$ is a

304: "magnetic" or an ordering field conjugate to $\chi$.  Because $\Delta

305: \chi$ is dimensionless, we expect that the ordering field $h \sim T$ and

306: thus $Td<\chi>/dT$ in proteins is similar to magnetic susceptibility. (2)

307: Camacho and Thirumalai  \cite{Camacho93PNAS} have suggested that

308: efficient folding in apparent two-state folders requires $T_F\approx

309: T_\theta$. Because the transition at $T_\theta$ is usually second order

310: \cite{GrosbergBook}, while the one at $T_F$ is first order [2c,11],

311: %\cite{Dill95ProtSci,Shakhnovich89Biop}, 

312: the $T_F\approx T_\theta$ condition implies that folding of two-state

313: globular proteins occurs near a tricritical point \cite{Camacho93PNAS}.  

314: Therefore, the critical exponents that control the

315: behavior of the polypeptide chain at $T_{\theta}$ should manifest itself

316: in the {\bf U} $\leftrightarrow$ {\bf NBA} phase transition. Using these

317: arguments we can obtain the $N$ dependence of $\Omega_c$ in the following

318: way. In general, we expect that close to $T \approx T_{\theta} \approx

319: T_F$ the Flory radius \cite{DeGennesBook} $R_F \sim \Delta T^{-\nu} \sim

320: N^{\nu}$ ($R_F$ is the analogue of the correlation length in magnetic

321: systems). This implies that $\Delta T/T_F \sim N^{-1}$. Because of the

322: analogy to magnetic susceptibility, we expect $Td<\chi>/dT \sim N^\gamma$.

323: Using Eq. (\ref{coop_lattice_eq}) we obtain the expected relationship

324: $\Omega_c \sim N^{1+\gamma}$, which directly follows from the 

325: hypothesis that $T_F\approx T_\theta$ for efficient two-state folders

326: \cite{Camacho93PNAS}. 

327: 

328: 

329: The scaling $\Omega_c \sim N^{\zeta}$ with $\zeta$ clearly different

330: from 2 may appear to be at odds with the idea that the structures in 

331: the NBA are sequence-specific. However, the global characteristics embodied

332: in the growth of $\Omega_c$ with $N$ are {\em valid only at} $T\approx T_F$. In

333: the neighborhood of this temperature the general characteristics

334: of the {\bf U} $\leftrightarrow$ {\bf NBA} transition are governed

335: by the properties of the unfolded states as $T_F$ is approached from above.

336: It has been shown that in the denaturated states ($T>T_F$) the global

337: properties like the gyration radius $R_g \sim N^{\nu}$ with

338: $\nu \approx 0.59$ {\em as expected for homopolymers} \cite{Kohn04}.

339: Similarly, the homopolymeric 

340: nature around $T_F$ is reflected in the growth of $\Omega_c$ with $N$. 

341: 

342: The finding that the folding transition

343: at $T_F$ occurs at a tricritical point suggests

344: that the native states of natural proteins are only marginally stable.

345: Because biological functions require transitions between different states,

346: it is logical to postulate that natural foldable proteins have evolved to

347: ensure $T_\theta \approx T_F$. The coil-globular transition at

348: $T_\theta$ is likely to be a second order transition involving no

349: discontinuity in the free energy. At $T_F$ the transition is of the

350: first order. The closeness of $T_F$ and $T_\theta$ implies that the

351: discontinuity of the free energy at $T_F$ cannot be large. As a result

352: the folded state is expected to be only marginally stable with respect

353: to the ensemble of denatured states. As argued elsewhere

354: \cite{Thirum95} this

355: condition is also equivalent to maximizing the ratio $T_F/T_g$, where

356: $T_g$ is a glass transition temperature \cite{Onuchic97}. 

357: Marginality condition may also

358: be a requirement for robustness of the folded state. This may explain why

359: small single domain proteins can tolerate a large number of mutations

360: without substantial changes in the native state.

361: It is also likely, as recently shown, that evolution has led to marginally

362: stable proteins that have maximum sequence-structure 

363: compatibility \cite{Xia02,Taverna02}.

364: 

365: 

366: 

367: 

368: This work was

369: supported in part by a KBN grant  No 1P03B01827

370:  and the National Science Foundation grant

371: (NSF CHE-0209340). We are grateful to R.B. Dyer, A.R. Fersht, and

372: N. Ferguson for providing us with unpublished thermal denaturation data. 

373: Mai Suan Li wishes to thank the hospitality of

374: IPST, Maryland, USA and ICTP, Trieste, where part of this work was done. 

375:  

376: 

377: \begin{references}

378: 

379: \bibitem{Poland} (a) D. Poland and H. A. Scheraga,

380: {\em Theory of helix-coil transitions in biopolymers}

381: (Academic Press, New York, 1970); (b) 

382: T. E. Creighton, {\em Proteins: Structures and Molecular Principles} 

383: (W. H. Freeman \& Co., New York, 1993); (c) P.L. Privalov, 

384: Adv. Phys. Chem. {\bf 33}, 167 (1979). 

385: 

386: \bibitem{Bryngelson95Proteins} (a) J. D. Bryngelson, J. N. Onuchic,  

387: N. D. Socci, and P. G. Wolynes,  Proteins Struct Funct Gen {\bf 21},

388: 167 (1995); (b) A. Kolinski, W. Galazka, and 

389: J. Skolnick,   Proteins: Struct Funct Gen {\bf 26}, 271 (1996); (c)

390: K. A. Dill {\em et al.}, Prot. Sci. {\bf 4}, 561 (1995).

391: 

392: %\bibitem{Skolnick96Proteins}  A. Kolinski, W. Galazka, and 

393: %J. Skolnick,   Proteins Struct Funct Gen {\bf 26}, 271 (1996). 

394: %\bibitem{Dill95ProtSci} K. A. Dill, S. Bromberg, K. Yue, K. M. Fiebig, 

395: %D. P. Yee, P. D. Thomas, and H. S. Chan. Protein Science {\bf 4}, 561

396: %(1995). 

397: %\bibitem{Dill95ProtSci} K. A. Dill {\em et al.}, 

398: %Protein Science {\bf 4}, 561 (1995).

399: 

400: 

401: 

402: \bibitem{Camacho93PNAS} C. J. Camacho and D. Thirumalai, 

403: Proc. Natl. Acad. Sci. USA {\bf 90},  6369 (1993). 

404: 

405: 

406: \bibitem{Scheraga98JMB}  M.-H. Hao and H. Scheraga,   

407: J Mol Biol {\bf 277}, 973 (1998). 

408: 

409: \bibitem{KlimThirum98FD} D. Klimov and D. Thirumalai, Fold. Des. {\bf

410: 3}, 127  (1998).  

411: 

412: 

413: \bibitem{KlimThirum02JCC} D. K. Klimov and D. Thirumalai, 

414: J. Comp. Chem. {\bf  23}, 161 (2002); M. S. Li,  D. K. Klimov and

415: D. Thirumalai, Polymer {\bf 45}, 573 (2004). 

416: 

417: 

418: \bibitem{Fisher} (a) M. E. Fisher and M. N. Barber, Phys. Rev. Lett. {\bf

419: 28}, 1516 (1972); (b)  M. E. Fisher and A. N. Berker, Phys. Rev. B

420: {\bf 26}, 2507 (1982). 

421: 

422: 

423: 

424: 

425: \bibitem{Thirum95} D. Thirumalai, J Phys I (France) {\bf 5},  1457 (1995). 

426: 

427: \bibitem{Levitt97ProtSci} (a) J. Tsai, M. Gerstein, and

428: M. Levitt, Prot. Sci. {\bf 6}, 2606 (1997); (b)  H. Kaya and

429: H.S. Chan, Proteins: Struct. Funct. Gen. {\bf 40}, 637 (2000); 

430: (c) H. Kaya and H.S. Chan, J. Mol. Biol. {\bf 325}, 911 (2003). 

431: 

432: 

433: \bibitem{LMCalculations}

434: We generated between 12 to 20 distinct maximally compact native

435: structures for a given value of $N$ ($27\le N\le 80$).  For example,

436: all native structures for $N=80$ were confined to  the vertices of a

437: 4x4x5 cube. In the Go model each structure corresponds to a

438: distinct sequence. We used, depending on $N$, between 50 and 100 Monte

439: Carlo trajectories to collect states and applied multiple histogram method

440: \cite{Ferrenberg} to compute $<\chi(T)>$,  $d<\chi>/dT$,  and 

441: $\Omega_c$ using Eq. (5). For the Go models (and other 

442: highly optimized sequences) $T_F$, identified with the location of the

443: maximum in $d<\chi>/dT$, coincides with $T_\theta$

444: \cite{Camacho93PNAS}. 

445: 

446: \bibitem{GrosbergBook} A. Yu. Grosberg and A. R. Khokhlov, 

447: {\em Statistical Physics of Macromolecules}  (AIP Press, New York, 1994). 

448: 

449: \bibitem{Shakhnovich89Biop} E.  Shakhnovich and A. Finkelstein,   

450: Biopolymers {\bf 28}, 1667 (1989). 

451: 

452: 

453: 

454: \bibitem{DeGennesBook} P. G. de Gennes, 

455: {\em Scaling Concepts in Polymer Physics} (Cornell University Press, 

456: New York, 1985). 

457: 

458: 

459: \bibitem{Ferrenberg} A. M.  Ferrenberg and  R. H. Swendsen, 

460: Phys. Rev. Lett. {\bf 63 }, 1195 (1989). 

461: 

462: 

463: \bibitem{ProteinDB} A  list of WT proteins together with the values of

464: parameters for thermal and chemical denaturation  is available at

465: {\small www.biotheory.umd.edu/ScalingDB.html} or 

466: {\small binf.gmu.edu/dklimov/Publications/ScalingDB.html}. In both 

467: datasets the list of proteins spans the whole range of topologies,

468: from peptides with $\alpha$-helical or $\beta$-hairpin

469: structures to all $\alpha$, all $\beta$, or $\alpha/\beta$

470: proteins. Because of spectrum of topologies there is very little

471: sequence similarity. The larger error in $\zeta$ and $\lambda$ for

472: chemical denaturation data is due to the shorter span of $N$. 

473: 

474: 

475: \bibitem{Kleinert} H. Kleinert and V. Schulte-Flohlinde, {\em Critical

476: Properties of $\phi^4$-Theories} (World Scientific, Singapore, 2002).

477: 

478: \bibitem{Kohn04} J. E. Kohn {\em et al.}, 

479: Proc. Natl. Acad. Sci. USA {\bf 101}, 12491 (2004).

480: 

481: \bibitem{Onuchic97} J. N. Onuchic, Z. A. Luthey-Schulten, and

482: P. G. Wolynes, Ann. Rev. Phys. Chem {\bf 48}, 545-600 (1997). 

483: 

484: \bibitem{Xia02} Y. Xia and M. Levitt, Proc. Natl. Acad. Sci. USA {\bf 99},

485: 10382 (2002).

486: 

487: \bibitem{Taverna02} D. M. Taverna and R. Goldstein, 

488: J. Mol. Biol. {\bf 315}, 49 (2002).

489: 

490: 

491: \end{references}

492: 

493: \end{document}

494: