We introduce an approach for computing the free energy and the probability density in high-dimensional spaces, such as those explored in molecular dynamics simulations of biomolecules. The approach exploits the presence of correlations between the coordinates, induced, in molecular dynamics, by the chemical nature of the molecules. Due to these correlations, the data points lay on a manifold that can be highly curved and twisted, but whose dimension is normally small. We estimate the free energies by finding, with a statistical test, the largest neighborhood in which the free energy in the embedding manifold can be considered constant. Importantly, this procedure does not require defining explicitly the manifold and provides an estimate of the error that is approximately unbiased up to large dimensions. We test this approach on artificial and real data sets, demonstrating that the free energy estimates are reliable for data sets on manifolds of dimension up to ∼10, embedded in an arbitrarily large space. In practical applications our method permits the estimation of the free energy in a space of reduced dimensionality without specifying the collective variables defining this space.

Computing the Free Energy without Collective Variables

Rodriguez, Alex;D'Errico, Maria;Laio, Alessandro
2018-01-01

Abstract

We introduce an approach for computing the free energy and the probability density in high-dimensional spaces, such as those explored in molecular dynamics simulations of biomolecules. The approach exploits the presence of correlations between the coordinates, induced, in molecular dynamics, by the chemical nature of the molecules. Due to these correlations, the data points lay on a manifold that can be highly curved and twisted, but whose dimension is normally small. We estimate the free energies by finding, with a statistical test, the largest neighborhood in which the free energy in the embedding manifold can be considered constant. Importantly, this procedure does not require defining explicitly the manifold and provides an estimate of the error that is approximately unbiased up to large dimensions. We test this approach on artificial and real data sets, demonstrating that the free energy estimates are reliable for data sets on manifolds of dimension up to ∼10, embedded in an arbitrarily large space. In practical applications our method permits the estimation of the free energy in a space of reduced dimensionality without specifying the collective variables defining this space.
2018
https://pubs.acs.org/doi/10.1021/acs.jctc.7b00916
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11368/3032523
 Avviso

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 33
  • ???jsp.display-item.citation.isi??? 33
social impact