Difference between revisions of "Simulation Acceleration"

From AlchemistryWiki
Jump to navigation Jump to search
m
Line 4: Line 4:
  
 
=Umbrella Sampling=
 
=Umbrella Sampling=
One standard method for improving sampling in atomistic simulations is umbrella sampling,{{cite|Torrie1977|Torrie, G. M., and Valleau, J. P. (1977) Non-physical Sampling Distributions in Monte-Carlo Free-Energy Estimation : Umbrella Sampling. ''J. Comput. Phys.'' 23, 187–199.|http://www.citeulike.org/group/14929/article/9052620}} where bias terms are added to constrain the simulation in some way, then the restraints are removed. This methods can lower barriers on the potential energy surface, or restrain simulations to slow-interconverting configurations that are relevant to the binding affinities (e.g. different torsional states). Doing this allows for the free energy components to be properly computed and then combined.
+
One standard method for improving sampling in atomistic simulations is umbrella sampling,{{cite|Torrie1977|Torrie, G. M., and Valleau, J. P. (1977) Non-physical Sampling Distributions in Monte-Carlo Free-Energy Estimation : Umbrella Sampling. ''J. Comput. Phys.'' 23, 187–199.|http://www.citeulike.org/group/14929/article/9052620}} where bias terms are added to constrain the simulation in some way, then the restraints are removed. This method can lower barriers on the potential energy surface, or restrain simulations to slow-interconverting configurations that are relevant to the binding affinities (e.g. different torsional states). Doing this allows for the free energy components to be properly computed and then combined.
 
{{cite|Wang2006|Wang, J., Deng, Y., and Roux, B. (2006) Absolute Binding Free Energy Calculations Using Molecular Dynamics Simulations with Restraining Potentials. ''Biophys. J.'' 91, 2798–2814.|http://www.citeulike.org/user/kupopo/article/867499}}<!--
 
{{cite|Wang2006|Wang, J., Deng, Y., and Roux, B. (2006) Absolute Binding Free Energy Calculations Using Molecular Dynamics Simulations with Restraining Potentials. ''Biophys. J.'' 91, 2798–2814.|http://www.citeulike.org/user/kupopo/article/867499}}<!--
 
-->{{cite|Mobley2007|Mobley, D. L., Chodera, J. D., and Dill, K. A. (2007) Confine-and-release method: Obtaining correct binding free energies in the presence of protein conformational change. ''J. Chem. Theory Comput.'' 3, 1231–1235.|http://www.citeulike.org/group/14929/article/9052442}}<!--
 
-->{{cite|Mobley2007|Mobley, D. L., Chodera, J. D., and Dill, K. A. (2007) Confine-and-release method: Obtaining correct binding free energies in the presence of protein conformational change. ''J. Chem. Theory Comput.'' 3, 1231–1235.|http://www.citeulike.org/group/14929/article/9052442}}<!--
Line 30: Line 30:
  
 
=Other Methods=
 
=Other Methods=
The other methods that will be mentioned briefly are ''Expanded Ensemble'' and ''<math>\lambda</math>-dynamics.'' Expanded ensemble works by running a single simulation and sampling both the intermediate states and separate coordinates simultaneously. <math>\lambda</math>-dynamics works by treating the alchemical variable as dynamical, introducing a fictitious mass corresponding to the \lambda degree of freedom; although this seems rather novel, it is essentially identical to Monte Carlo techniques.
+
The other methods that will be mentioned briefly are ''Expanded Ensemble'' and ''<math>\lambda</math>-dynamics.'' Expanded ensemble works by running a single simulation and sampling both the intermediate states and separate coordinates simultaneously. <math>\lambda</math>-dynamics works by treating the alchemical variable as dynamical, introducing a fictitious mass corresponding to the <math>\lambda</math> degree of freedom; although this seems rather novel, it is essentially identical to Monte Carlo techniques.
 
{{cite|Banba2000}}<!--
 
{{cite|Banba2000}}<!--
 
-->{{cite|Guo1998|Guo, Z., Brooks III, C. L., and Kong, X. (1998) Efficient and flexible algorithm for free energy calculations using the <math>\lambda</math>-dynamics approach. ''J. Phys. Chem. B'' 102, 2032–2036.|http://www.citeulike.org/group/14929/article/9052235}}<!--
 
-->{{cite|Guo1998|Guo, Z., Brooks III, C. L., and Kong, X. (1998) Efficient and flexible algorithm for free energy calculations using the <math>\lambda</math>-dynamics approach. ''J. Phys. Chem. B'' 102, 2032–2036.|http://www.citeulike.org/group/14929/article/9052235}}<!--

Revision as of 15:18, 8 January 2013



In many cases of interest, carrying out robust free energy calculations may require significant investment of computational resources, beyond that which can be obtained by most researchers. This page examines additional tools for accelerating the sampling. As these begin to branch out from the fundamentals of free energy, we will not go deeply into all these methods. They are not needed to carry out free energy calculations, but they may be extremely useful to converge calculations in complex systems with slow dynamics and we encourage readers to examine these topics to see what will work for them.

Umbrella Sampling

One standard method for improving sampling in atomistic simulations is umbrella sampling,[1] where bias terms are added to constrain the simulation in some way, then the restraints are removed. This method can lower barriers on the potential energy surface, or restrain simulations to slow-interconverting configurations that are relevant to the binding affinities (e.g. different torsional states). Doing this allows for the free energy components to be properly computed and then combined. [2][3][3] If a system has slow degrees of freedom, like some hydration free energies,[4] this method allows one to sample more frequently in the slow state.

One can also envision using umbrella sampling to compute the free energy of constraining the free ligand into the bound conformation directly before computing the free energy of binding, and then computing the free energy of releasing these restraints in solution. Doing so usually decreases correlation times for sampling intermediate states, thereby increasing the simulation efficiency. [2][5]

Hamiltonian Replica Exchange

Although it is perfectly valid to run each intermediate state as its own simulation, there are tools that let you run parallel, coupled simulations and swap information between them to improve sampling. One such method is called Hamiltonian Exchange (HEX) or Hamiltonian Replica Exchange (HREX). In HEX, the parallel simulations at each alchemical variable are allowed to swap atoms/molecules (under certain conditions) from state A which has different energy barriers than state B. This enables sampling configurations that may take significant amounts of time to access normally when simulating only one state. [6][7][8][9][10][11]

We highly recommend doing HEX methods as most codes do this on top of well-tested temperature replica exchange methods, and so there is very little overhead. the outputs of Hamiltonian exchange simulations can be analyzed in the same way as the outputs of [math]\displaystyle{ K }[/math] uncoupled simulations. These simulations are guaranteed to decorrelate as fast or faster than standard simulations, though the exact amount of improvement depends on the system; at least they will not be any worse!

Correlation times with HEX is a bit more complicated since atoms/molecules are swapping along very different trajectories. In this case, you will need to compute along trajectories that sample different states as opposed to a single state.

Other Methods

The other methods that will be mentioned briefly are Expanded Ensemble and [math]\displaystyle{ \lambda }[/math]-dynamics. Expanded ensemble works by running a single simulation and sampling both the intermediate states and separate coordinates simultaneously. [math]\displaystyle{ \lambda }[/math]-dynamics works by treating the alchemical variable as dynamical, introducing a fictitious mass corresponding to the [math]\displaystyle{ \lambda }[/math] degree of freedom; although this seems rather novel, it is essentially identical to Monte Carlo techniques. [9][12][13][14] [math]\displaystyle{ \lambda }[/math]-dynamics is showing promise, but is still in preliminary stages of development.[15][16][17][14][18]

In general, we do NOT recommend Expanded Ensemble or [math]\displaystyle{ \lambda }[/math]-dynamics to beginners. The methods and implementations are not up to the same robustness as HEX yet and there are tweaks and extra parameters that have to be coded to obtain proper convergence. Given more time and development, these methods may become more accessible to beginners and we will be able to recommend them in the future.

References

  1. Torrie, G. M., and Valleau, J. P. (1977) Non-physical Sampling Distributions in Monte-Carlo Free-Energy Estimation : Umbrella Sampling. J. Comput. Phys. 23, 187–199. - Find at Cite-U-Like
  2. 2.0 2.1 Wang, J., Deng, Y., and Roux, B. (2006) Absolute Binding Free Energy Calculations Using Molecular Dynamics Simulations with Restraining Potentials. Biophys. J. 91, 2798–2814. - Find at Cite-U-Like
  3. 3.0 3.1 Mobley, D. L., Chodera, J. D., and Dill, K. A. (2007) Confine-and-release method: Obtaining correct binding free energies in the presence of protein conformational change. J. Chem. Theory Comput. 3, 1231–1235. - Find at Cite-U-Like Cite error: Invalid <ref> tag; name "Mobley2007" defined multiple times with different content
  4. Klimovich, P. V., and Mobley, D. L. (2010) Predicting hydration free energies using all-atom molecular dynamics simulations and multiple starting conformations. J. Comp. Aided Mol. Design 24, 307–316. - Find at Cite-U-Like
  5. Woo, H.-J., and Roux, B. (2005) Calculation of absolute protein-ligand binding free energy from computer simulation. Proc. Natl. Acad. Sci. 102, 6825–6830. - Find at Cite-U-Like
  6. Okamoto, Y. (2004) Generalized-ensemble algorithms: Enhanced sampling techniques for Monte Carlo and molecular dynamics simulations. J. Mol. Graph. Model. 22, 425–439. - Find at Cite-U-Like
  7. Roux, B., and Faraldo-Gómez, J. D. (2007) Characterization of conformational equilibria through Hamiltonian and temperature replica- exchange simulations: Assessing entropic and environmental effects. J. Comput. Chem. 28, 1634–1647. - Find at Cite-U-Like
  8. Woods, C. J., Essex, J. W., and King, M. A. (2003) Enhanced Configurational Sampling in Binding Free Energy Calculations. J. Phys. Chem. B 107, 13711–13718. - Find at Cite-U-Like
  9. 9.0 9.1 Banba, S., Guo, Z., and Brooks III, C. L. (2000) Efficient sampling of ligand orientations and conformations in free energy calculations using the lambda-dynamics method. J. Phys. Chem. B 104, 6903–6910. - Find at Cite-U-Like
  10. Bitetti-Putzer, R., Yang, W., and Karplus, M. (2003) Generalized ensembles serve to improve the convergence of free energy simulations. Chem. Phys. Lett. 377, 633–641. - Find at Cite-U-Like
  11. Hritz, J., and Oostenbrink, C. (2008) Hamiltonian replica exchange molecular dynamics using soft-core interactions. J. Chem. Phys. 128, 144121. - Find at Cite-U-Like
  12. Guo, Z., Brooks III, C. L., and Kong, X. (1998) Efficient and flexible algorithm for free energy calculations using the [math]\displaystyle{ \lambda }[/math]-dynamics approach. J. Phys. Chem. B 102, 2032–2036. - Find at Cite-U-Like
  13. Kong, X., and Brooks III, C. L. (1996) [math]\displaystyle{ \lambda }[/math]-dynamics: A new approach to free energy calculations. J. Chem. Phys. 105, 2414–2423. - Find at Cite-U-Like
  14. 14.0 14.1 Li, H., Fajer, M., and Yang, W. (2007) Simulated scaling method for localized enhanced sampling and simultaneous "alchemical" free energy simulations: A general method for molecular mechanical, quantum mechanical, and quantum mechanical/molecular mechanical simulations. J. Chem. Phys. 126, 024106. - Find at Cite-U-Like Cite error: Invalid <ref> tag; name "Li2007" defined multiple times with different content
  15. Zheng, L., Carbone, I. O., Lugovskoy, A., Berg, B. A., and Yang, W. (2008) A hybrid recursion method to robustly ensure convergence efficiencies in the simulated scaling based free energy simulations. J. Chem. Phys. 129, 034105. - Find at Cite-U-Like
  16. Zheng, L., and Yang, W. (2008) Essential energy space random walks to accelerate molecular dynamics simulations: Convergence improvements via an adaptive-length selfhealing strategy. J. Chem. Phys. 129, 014105.
  17. Min, D., and Yang, W. (2008) Energy difference space random walk to achieve fast free energy calculations. J. Chem. Phys. 128, 191102.
  18. Min, D., Li, H., Li, G., Bitetti-Putzer, R., and Yang, W. (2007) Synergistic approach to improve “alchemical” free energy calculation in rugged energy surface. J. Chem. Phys. 126, 144109. - Find at Cite-U-Like