Skip to main content
Log in

A minimization principle for transition paths of maximum flux for collective variables

  • Regular Article
  • Published:
Theoretical Chemistry Accounts Aims and scope Submit manuscript

Abstract

Considered is the construction of transition paths of conformational changes for proteins and other macromolecules, using methods that do not require the generation of dynamics trajectories. Special attention is given to the use of a reduced set of collective variables for describing such paths. A favored way to define transition paths is to seek channels through the transition state having cross sections with a high reactive flux (density of last hitting points of reactive trajectories). Given here is a formula for reactive flux that is independent of the parameterization of “collective variable space.” This formula is needed for the principal curve of the reactive flux (as in the revised finite temperature string method) and for the maximum flux transition (MaxFlux) path. Additionally, a resistance functional is derived for narrow tubes, which when minimized yields a MaxFlux path. A strategy for minimization is outlined in the spirit of the string method. Finally, alternative approaches based on determining trajectories of high probability are considered, and it is observed that they yield paths that depend on the parameterization of collective variable space, except in the case of zero temperature, where such a path coincides with a MaxFlux path.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

References

  1. Berkowitz M, Morgan JD, McCammon JA, Northrup SH (1983) Diffusion-controlled reactions: a variational formula for the optimum reaction coordinate. J Chem Phys 79:5563–5565

    Article  CAS  Google Scholar 

  2. Bolhuis PG, Chandler D, Dellago C, Geissler PL (2002) Transition path sampling: throwing ropes over rough mountain passes, in the dark. Annu Rev Phys Chem 53:291–318

    Article  CAS  Google Scholar 

  3. Burger SK, Yang W (2006) Quadratic string method for determining the minimum-energy path based on multiobjective optimization. J Chem Phys 124:054109-1–054109-13

    Google Scholar 

  4. Cameron MK (2013) Estimation of reactive fluxes in gradient stochastic systems using an analogy with electric circuits. J Comput Phys 247:137–152

    Article  CAS  Google Scholar 

  5. Weinan E, Ren W, Vanden-Eijnden E (2002) String method for the study of rare events. Phys Rev B 66:6688–6693

    Google Scholar 

  6. Weinan E, Ren W, Vanden-Eijnden E (2004) Minimum action method for the study of rare events. Commun Pure Appl Math 57(5):637–656

    Article  Google Scholar 

  7. Weinan E, Ren W, Vanden-Eijnden E (2005) Finite temperature string method for the study of rare events. J Phys Chem B 109(5):6688–6693

    Google Scholar 

  8. Weinan E, Ren W, Vanden-Eijnden E (2007) Simplified and improved string method for computing the minimum energy paths in barrier-crossing events. J Chem Phys 126:164103-1–164103-8

    Google Scholar 

  9. Weinan E, Vanden-Eijnden E (2006) Towards a theory of transition paths. J Stat Phys 123(3):503–523

    Article  Google Scholar 

  10. Weinan E, Vanden-Eijnden E (2010) Transition-path theory and path-finding algorithms for the study of rare events. Annu Rev Phys Chem 61(1):391–420

    Article  Google Scholar 

  11. Faccioli P, Sega M, Pederiva F, Orland H (2006) Dominant pathways in protein folding. Phys Rev Lett 97:108101

    Article  CAS  Google Scholar 

  12. Huang H, Zhao R, Dickson BM, Skeel RD, Post CB (2012) \(\alpha\)C helix as a switch in the conformational transition of Src/CDK-like kinase domains. J Phys Chem B 116(15):4465–4475. doi:10.1021/jp301628r

    Article  CAS  Google Scholar 

  13. Huo S, Straub JE (1997) The MaxFlux algorithm for calculating variationally optimized reaction paths for conformational transitions in many body system at finite temperature. J Chem Phys 107(13):5000–5006

    Article  CAS  Google Scholar 

  14. Jiménez A, Crehuet R (2007) Can the MaxFlux algorithm describe bifurcating paths? Theor Chem Acc 118:769–775

    Article  Google Scholar 

  15. Johnson ME, Hummer G (2012) Characterization of a dynamic string method for the construction of transition pathways in molecular reactions. J Phys Chem B 116:8573–8583

    Article  CAS  Google Scholar 

  16. Maragliano L, Fischer A, Vanden-Eijnden E, Ciccotti G (2006) String method in collective variables: minimum free energy paths and isocommittor surfaces. J Chem Phys 125(8):024106-1–024106-15

    Google Scholar 

  17. Maragliano L, Roux B, Vanden-Eijnden E (2014) Comparison between mean forces and swarms-of-trajectories string methods. J Chem Theory Comput 10(2):524–533

    Article  CAS  Google Scholar 

  18. Metzner P, Schütte C, Vanden-Eijnden E (2006) Illustration of transition path theory on a collection of simple examples. J Chem Phys 125(8):084110-1–084110-17

    Article  Google Scholar 

  19. Olender R, Elber R (1997) Yet another look at the steepest descent path. J Mol Struct THEOCHEM 398–399:63–71

    Article  Google Scholar 

  20. Pan AC, Sezer D, Roux B (2008) Finding transition pathways using the string method with swarms of trajectories. J Phys Chem B 112(11):3432–3440

    Article  CAS  Google Scholar 

  21. Park S, Sener MK, Lu D, Schulten K (2003) Reaction paths based on mean first-passage times. J Chem Phys 119(3):1313–1319

    Article  CAS  Google Scholar 

  22. Singhal N, Pande VS (2005) Error analysis and efficient sampling in Markovian state models for molecular dynamics. J Chem Phys 123(20):4909

    Article  Google Scholar 

  23. Vanden-Eijnden E (2006) Transition path theory. In: Binder K, Ciccotti G, Ferrario M (eds) Computer simulations in condensed matter: from materials to chemical biology, vol 2., Lecture notes in physicsSpringer, Berlin, pp 453–493

    Google Scholar 

  24. Vanden-Eijnden E, Heymann M (2008) The geometric minimum action method for computing minimum energy paths. J Chem Phys 128(6):61103

    Article  Google Scholar 

  25. West AM, Elber R, Shalloway D (2007) Extending molecular dynamics timescales with milestoning: example of complex kinetics in a solvated peptide. J Chem Phys 126:145104

    Article  Google Scholar 

  26. Wolynes P (1989) Chemical reaction dynamics in complex molecular systems. In: Stein DL (ed) Lectures in the sciences of complexity: the proceedings of the 1988 complex systems summer school held June–July, 1988 in Santa Fe, New Mexico, Santa Fe institute studies in the sciences of complexity lectures. Addison-Wesley, Reading

    Google Scholar 

  27. Zhao R, Shen J, Skeel RD (2010) Maximum flux transition paths of conformational change. J Chem Theory Comput 6(8):2411–2423

    Article  CAS  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Robert D. Skeel.

Additional information

This material is based upon work supported by Grant R01GM083605 from the National Institute of General Medical Sciences.

Appendix

Appendix

1.1 Area metric

A formula is derived for measuring surface area with a metric tensor \(G(\zeta )\). An infinitesimal surface \(\delta \varSigma\) enclosing a point \(\zeta _0\) can be expressed as the intersection of an infinitesimal region \(\delta \varOmega\) surrounding \(\zeta _0\) and the plane \(\hat{n}\cdot (\zeta -\zeta _0) = 0\). To determine the area of \(\delta \varSigma\), introduce variables \(\zeta ' = G^{1/2}(\zeta -\zeta _0)\), for which distances and angles are locally Euclidean. In the new coordinates, the surface becomes the intersection of some transformed region \(\delta \varOmega '\) and the plane \(\hat{n}'\cdot \zeta ' = 0\) where \(\hat{n}' = G^{-1/2}\hat{n}/|G^{-1/2}\hat{n}|\) and \(|{\delta \zeta }| = ({\delta \zeta }^{\mathsf {T}}{\delta \zeta })^{1/2}\). Its surface area is

$$\begin{aligned} \int _{\delta \varSigma '}{\mathrm {d}} S_{\zeta '}= & {} \int _{\delta \varOmega '}\delta (\hat{n}'\cdot \zeta '){\mathrm {d}}\zeta ' \\= & {} \int _{\delta \varOmega } \delta (\hat{n}'\cdot G^{1/2}(\zeta -\zeta _0))\det G^{1/2} {\mathrm {d}}\zeta \\= & {} \int _{\delta \varOmega } \delta ((\hat{n}/|G^{-1/2}\hat{n}|)\cdot (\zeta -\zeta _0))(\det G)^{1/2} {\mathrm {d}}\zeta \\= & {} \int _{\delta \varOmega } |G^{-1/2}\hat{n}|\delta (\hat{n}\cdot (\zeta -\zeta _0))(\det G)^{1/2} {\mathrm {d}}\zeta \\= & {} \int _{\delta \varSigma }|G^{-1/2}\hat{n}|(\det G)^{1/2}{\mathrm {d}} S_\zeta \end{aligned}$$

where the first and last equality follow by applying the coarea formula This result is restated in Eq. (8) using the collective variable metric.

1.2 Path of maximum reactive flux

Define \({\hat{q}}(s) = q(Z(s))\), and define \(\sigma (\zeta )\) implicitly by \(q(\zeta ) = {\hat{q}}(\sigma (\zeta ))\). The assumption Eq. (10) that isocommittors are planar implies that \(\sigma (\zeta )\) satisfies

$$\begin{aligned} \hat{n}(\sigma (\zeta ))^{\mathsf {T}}(\zeta - Z(\sigma (\zeta ))) = 0, \end{aligned}$$

so on each isocommittor \(\varPi (s)\),

$$\begin{aligned} \nabla \sigma (\zeta )|_{\sigma (\zeta ) = s} = \frac{ \hat{n}(s)}{Z_s(s)^{\mathsf {T}}\hat{n}(s) -\hat{n}_s(s)^{\mathsf {T}}(\zeta - Z(s))}. \end{aligned}$$
(29)

 It is convenient to use Eq. (12) to write Eq. (9) as

$$\begin{aligned} j(\zeta ) = d_0 C_\xi \gamma (\zeta )^{-1}c(\zeta , G(\zeta )^{-1}\nabla q(\zeta )). \end{aligned}$$
(30)

Using Eq. (29) and \(q(\zeta ) = {\hat{q}}(\sigma (\zeta ))\), Eq. (30) becomes

$$\begin{aligned} j(\zeta )|_{\sigma (\zeta ) = s} = d_0 C_\xi {\hat{q}}_s(s) \frac{\gamma (\zeta )^{-1}c(\zeta , G(\zeta )^{-1}\hat{n}(s))}{\hat{n}(s)^{\mathsf {T}}Z_s(s) -\hat{n}_s(s)^{\mathsf {T}}(\zeta - Z(s))}. \end{aligned}$$

Maximizing \(j(\zeta )\) over \(\zeta\) subject to \(\hat{n}(s)^{\mathsf {T}}(\zeta - Z(s)) = 0\) gives the condition

$$\begin{aligned} -\frac{\nabla _\zeta \gamma (Z)}{\gamma (Z)} -\frac{(\nabla _\zeta c)(Z, G(Z)^{-1}\hat{n})}{c(Z, G(Z)^{-1}\hat{n})} +\frac{\hat{n}_s}{\hat{n}^{\mathsf {T}}Z_s}~\Vert ~\hat{n} \end{aligned}$$
(31)

for the value Z(s) that maximizes \(j(\zeta )\). (See Ref. [27], Sec. 3.4.)

The assumption that the flux J(Z) is parallel to the direction \(Z_s\) of the path implies \(G(Z)^{-1}\hat{n}~\Vert ~Z_s\). Noting that \(\nabla _\omega c(\zeta ,\omega ) = G(\zeta )\omega /c(\zeta ,\omega )\), condition (31) becomes

$$\begin{aligned} -\frac{\nabla _\zeta \gamma (Z)}{\gamma (Z)} -\frac{\nabla _\zeta c(Z, Z_s)}{c(Z, Z_s)} +\frac{(\nabla _\omega c(Z, Z_s))_s}{c(Z, Z_s)}~\Vert ~\nabla _\omega c(Z, Z_s). \end{aligned}$$

(See Ref. [27], Eq. (10).) This can be written as an equation by expressing the left-hand side as an undetermined scalar times the right-hand side and then solving for the scalar after premultiplying this equation by \(Z_s^{\mathsf {T}}\). The result is Eq. (11).

1.3 Narrow tube flow rate formula

Using Eq. (6) and Eq. (18) of Ref. [18], the flow rate can be expressed as a volume integral:

$$\begin{aligned} \text{ flow } \text{ rate } =\int _\varOmega \rho _\xi (\zeta )\nabla q(\zeta )^{\mathsf {T}}D(\zeta )\nabla q(\zeta ){\mathrm {d}}\zeta = C_\xi d_0 I(q). \end{aligned}$$
(32)

Foliating \(\varOmega\) in Eq. (6) gives

$$\begin{aligned}I(q) = \int _0^1 \int _\varOmega {\mathrm {e}}^{-\beta F(\zeta )} \nabla q(\zeta )^{\mathsf {T}}G(\zeta )^{-1}\nabla q(\zeta )\delta (\sigma (\zeta ) - s) {\mathrm {d}}\zeta \,{\mathrm {d}} s \end{aligned}$$

where the decomposition \(\hat{q}(\sigma (\zeta )) = q(\zeta )\) is defined by Eq. (13). Using the coarea formula gives

$$\begin{aligned} I(q) =\int _0^1\int _{\varSigma (s)} \frac{\mathrm {e}^{-\beta F(\zeta )}\nabla q(\zeta )^{\mathsf {T}}G(\zeta )^{-1}\nabla q(\zeta )}{|\nabla \sigma (\zeta )|} \mathrm {d}S_\zeta \,\mathrm {d}s. \end{aligned}$$

Using \(\nabla q(\zeta ) = {\hat{q}}_s(s)\nabla \sigma (\zeta )\) gives

$$\begin{aligned} I(q)= & {} \int _0^1{\hat{q}}_s(s)^2 \int _{\varSigma (s)}\mathrm {e}^{-\beta F(\zeta )} \nabla \sigma (\zeta )^{\mathsf {T}}G(\zeta )^{-1}\hat{n}(\zeta ) {\mathrm {d}}S_\zeta \,{\mathrm {d}}s \\= & {} \int _0^1{\hat{q}}_s(s)^2 Q(s){\mathrm {d}}s \end{aligned}$$

where \(\hat{n}(\zeta ) = \nabla \sigma (\zeta )/|\nabla \sigma (\zeta )|\) is the outer normal at \(\zeta\) on \(\varSigma (s)\) and Q(s) is given by Eq. (15). Minimizing over \({\hat{q}}\) for fixed \(\sigma\) gives the Euler–Lagrange equation \(-(2{\hat{q}}_s Q(s))_s = 0\), whose solution is given by

$$\begin{aligned} q(Z(s)) = {\hat{q}}(s) = I(q)\int _0^s\frac{\mathrm {d}s}{Q(s)}. \end{aligned}$$

Therefore,

$$\begin{aligned} I(q) = \left( \int _0^1 \frac{\mathrm {d}s}{Q(s)}\right) ^{-1}. \end{aligned}$$
(33)

From Eqs. (33) and (32), one obtains Eq. (14).

1.4 Direction of steepest descent

Let \((\int _0^1 \delta Z(s)^{\mathsf {T}}G(Z(s))\delta Z(s){\mathrm {d}}s)^{1/2}\) measure the size of a deviation \(\delta Z(s)\) from a path Z(s). For the direction of steepest descent f for the functional (17) subject to the constraint (23) the Lagrangian is

$$\begin{aligned} \int _0^1 f^{\mathsf {T}}h\,{\mathrm {d}}s +\mu \left( \int _0^1 f^{\mathsf {T}}G f\,\mathrm {d}s -{\mathrm {const}}\right) +\int _0^1\lambda ((f^{\mathsf {T}}b)_s + (f^{\mathsf {T}}\nabla _\omega c)_{ss}){\mathrm {d}}s. \end{aligned}$$

Its first variation is

$$\begin{aligned} \int _0^1 (\delta f)^{\mathsf {T}}(h -\lambda _s b +\lambda _{ss}\nabla _\omega c + 2\mu G f){\mathrm {d}}s + \left[ (\delta f)_s^{\mathsf {T}}\lambda \nabla _\omega c\right] _0^1. \end{aligned}$$

The Euler–Lagrange equations give \(\lambda (0) =\lambda (1) = 0\) and \(f = -(2\mu G)^{-1}(h -\lambda _s b +\lambda _{ss} \nabla _\omega c)\), in addition to the two original constraints. The value of \({\mathrm {const}}\) can be chosen to make \(\mu =\frac{1}{2}\), giving Eq. (25). With this choice, Eq. (23) becomes

$$\begin{aligned} (-\lambda _{sss} + b^{\mathsf {T}}G^{-1} b\lambda _s - b^{\mathsf {T}}G^{-1}h)_s = 0. \end{aligned}$$

where \(b^{\mathsf {T}}G^{-1}\nabla _\omega c = 0\) follows from Eq. (20). Also, using Eq. (20), one has for \(s = 0, 1\) that

$$\begin{aligned} 0 = Z_s^{\mathsf {T}}G f = -\lambda _{ss}Z_s^{\mathsf {T}}\nabla _\omega c =\lambda _{ss}c^2, \end{aligned}$$

thus obtaining two more boundary conditions for \(\lambda\), namely \(\lambda _{ss}(0) =\lambda _{ss}(1) = 0\).

1.5 Most probable path

As a first step in maximizing the action, given by Eq. (26), separate the path of the trajectory from its kinetics by a reparameterization \(t =\tau (s)\), \(0\le s\le 1\), and define the path by

$$\begin{aligned} Z(s) ={\hat{\zeta }}(\tau (s)). \end{aligned}$$

(For example, choose \(\tau (s)\) so that s is relative arclength, meaning that \(|Z_s(s)| = L\) where L is the length of the path.) With this change of variables, the action becomes

$$\begin{aligned}S[{\hat{\zeta }}] =\int _0^1\left| \tau _s^{-1}Z_s + f(Z)\right| _Z^2\tau _s\,{\mathrm {d}}s. \end{aligned}$$

The schedule \(t =\tau (s)\) that minimizes the action subject to \(\tau (0) = 0\) and \(\tau (1) = T\) is determined using the calculation of variations. The result is that \(\tau _s = |Z_s|_Z(|f|_Z^2 + C)^{-1/2}\) where C satisfies Eq. (28). With this choice, the action becomes \({\bar{S}}[Z]\), given in Eq. (27).

Consider now the effect of a change of variables on the functional. In particular, if one use instead variables \(\eta\), where

$$\begin{aligned} \zeta\,=\,\chi (\eta ), \end{aligned}$$

one has \(Z = \chi (\bar{Z})\) and \(\bar{F}(\eta )\), \(\bar{D}(\eta )\) given by

$$\begin{aligned} \bar{F}= & {} F(\chi ) -\frac{1}{\beta }\log \left| \det \frac{\partial \chi }{\partial \eta }\right| , \\ \bar{D}= & {} \left( \frac{\partial \chi }{\partial \eta }\right) ^{-1} D(\chi )\left( \frac{\partial \chi }{\partial \eta }\right) ^{-{\mathsf {T}}}, \end{aligned}$$

where the last two equations are derived from definitions given by Eqs. (2) and (3). Therefore, \(\bar{Z}_s = ((\partial \chi /\partial \eta )(\bar{Z}))^{-1}Z_s\) and \(\bar{s}[\bar{Z}] = \bar{S}[Z]\) provided \(\bar{f} =(\partial \chi /\partial \eta )^{-1}f(\chi )\) where

$$\begin{aligned} \bar{f} =\beta \bar{D}\nabla _\eta \bar{F} -{\mathrm {div}}_\eta \bar{D}. \end{aligned}$$

Differentiating the expression for \(\bar{F}\) gives

$$\begin{aligned} \left( \frac{\partial \chi }{\partial \eta }\right) ^{\mathsf {T}}(\nabla _\zeta F)(\chi ) =\nabla _\eta \left( \bar{F} +\frac{1}{\beta }\log \left| \det \frac{\partial \chi }{\partial \eta }\right| \right) , \end{aligned}$$

whence

$$\begin{aligned} f(\chi ) =\beta D(\chi )(\nabla _\zeta F)(\chi ) -({\mathrm {div}}D)(\chi ) = \frac{\partial \chi }{\partial \eta }\bar{f} + g, \end{aligned}$$

and

$$\begin{aligned} g =\frac{\partial \chi }{\partial \eta }\bar{D}\nabla _\eta \left( \log \left| \det \frac{\partial \chi }{\partial \eta }\right| \right) -({\mathrm {div}}D)(\chi ) +\frac{\partial \chi }{\partial \eta }{\mathrm {div}}\bar{D}. \end{aligned}$$

One has

$$\begin{aligned} \left( \frac{\partial \chi }{\partial \eta }\bar{D}\nabla \left( \log \left| \det \frac{\partial \chi }{\partial \eta }\right| \right) \right) ^i =\chi ^i_j\bar{D}^{jk}\left( \log \left| \det \frac{\partial \chi }{\partial \eta }\right| \right) _k =\chi ^i_j\bar{D}^{jk} {\mathrm {tr}}\left( \left( \frac{\partial \chi }{\partial \eta }\right) ^{-1}\left( \frac{\partial \chi }{\partial \eta }\right) _k\right) , \end{aligned}$$

where tensor notation is used, with subscripts for differentiation. Differentiating \(D^{ij}(\chi )\) gives

$$\begin{aligned} D^{ij}_k(\chi )\chi ^k_l = (\chi ^i_m\bar{D}^{mn}\chi _n^j)_l, \end{aligned}$$

whence

$$\begin{aligned} D^{ij}_k(\chi ) = (\chi ^i_m\bar{D}^{mn}\chi _n^j)_l\bar{\chi }^l_k, \end{aligned}$$

where \(\bar{\chi }^i_j\) is element ij of \((\partial \chi /\partial \eta )^{-1}\), so

$$\begin{aligned} ({\mathrm {div}}D)^i(\chi )=\, (\chi ^i_m\bar{D}^{mn}\chi ^j_n)_l\bar{\chi }^l_j \\=\, \chi ^i_{ml}\bar{D}^{ml} +\chi ^i_m\bar{D}^{ml}_l +\chi ^i_m\bar{D}^{mn}\chi ^j_{nl}\bar{\chi }^k_j. \end{aligned}$$

Therefore,

$$\begin{aligned} g^i ={\mathrm {tr}}\left( \bar{D}\frac{\partial ^2\chi ^i}{\partial \eta \partial \eta } \right) . \end{aligned}$$

The extra term g changes the form of the functional to be minimized, so if one minimizes the integral using instead variables \(\eta\), the resulting path \(\eta = \bar{Z}(s)\) does not satisfy \(Z(s) = \chi (\bar{Z}(s))\).

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Skeel, R.D., Zhao, R. & Post, C.B. A minimization principle for transition paths of maximum flux for collective variables. Theor Chem Acc 136, 14 (2017). https://doi.org/10.1007/s00214-016-2041-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s00214-016-2041-3

Keywords

Navigation