Markov state models from short non-equilibrium simulations—Analysis and correction of estimation bias

Nüske, Feliks; Wu, Hao; Prinz, Jan-Hendrik; Wehmeyer, Christoph; Clementi, Cecilia; Noé, Frank

doi:10.1063/1.4976518

Cited by 62 publications

(78 citation statements)

References 33 publications

(99 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A 250 nanosecond MD trajectory generated in Ref. [70] (MD setup described there) serves as a dataset. The solute coordinates were stored every picosecond, resulting in 250, 000 configurations that are all aligned on the first frame using minimal RMSD fit to remove global translation and rotation.…”

Section: G Alanine Dipeptidementioning

confidence: 99%

VAMPnets for deep learning of molecular kinetics

et al. 2018

Self Cite

View full text Add to dashboard Cite

There is an increasing demand for computing the relevant structures, equilibria, and long-timescale kinetics of biomolecular processes, such as protein-drug binding, from high-throughput molecular dynamics simulations. Current methods employ transformation of simulated coordinates into structural features, dimension reduction, clustering the dimension-reduced data, and estimation of a Markov state model or related model of the interconversion rates between molecular structures. This handcrafted approach demands a substantial amount of modeling expertise, as poor decisions at any step will lead to large modeling errors. Here we employ the variational approach for Markov processes (VAMP) to develop a deep learning framework for molecular kinetics using neural networks, dubbed VAMPnets. A VAMPnet encodes the entire mapping from molecular coordinates to Markov states, thus combining the whole data processing pipeline in a single end-to-end framework. Our method performs equally or better than state-of-the-art Markov modeling methods and provides easily interpretable few-state kinetic models.

show abstract

Section: G Alanine Dipeptidementioning

confidence: 99%

VAMPnets for deep learning of molecular kinetics

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…Alanine dipeptide is a representative model system for the characterization of conformational dynamics. Previous work [27,31,[50][51][52] has shown that the (ϕ, ψ) backbone dihedral angles act as ideal collective variables for describing the metastable configurational basins and associated transition kinetics, making it an excellent system for testing the GMVAE framework within a more realistic molecular simulation context. Since in general the optimal set of input features is unknown a priori, we use this example to test the ability of the GMVAE to identify the proper collective variables from a larger set of input features.…”

Section: Alanine Dipeptidementioning

confidence: 99%

Interpretable embeddings from molecular simulations using Gaussian mixture variational autoencoders

Varolgunes

Bereau

Rudzinski

2020

Mach. Learn.: Sci. Technol.

View full text Add to dashboard Cite

Extracting insight from the enormous quantity of data generated from molecular simulations requires the identification of a small number of collective variables whose corresponding low-dimensional free-energy landscape retains the essential features of the underlying system. Data-driven techniques provide a systematic route to constructing this landscape, without the need for extensive a priori intuition into the relevant driving forces. In particular, autoencoders are powerful tools for dimensionality reduction, as they naturally force an information bottleneck and, thereby, a low-dimensional embedding of the essential features. While variational autoencoders ensure continuity of the embedding by assuming a unimodal Gaussian prior, this is at odds with the multi-basin free-energy landscapes that typically arise from the identification of meaningful collective variables. In this work, we incorporate this physical intuition into the prior by employing a Gaussian mixture variational autoencoder (GMVAE), which encourages the separation of metastable states within the embedding. The GMVAE performs dimensionality reduction and clustering within a single unified framework, and is capable of identifying the inherent dimensionality of the input data, in terms of the number of Gaussians required to categorize the data. We illustrate our approach on two toy models, alanine dipeptide, and a challenging disordered peptide ensemble, demonstrating the enhanced clustering effect of the GMVAE prior compared to standard VAEs. The resulting embeddings appear to be promising representations for constructing Markov state models, highlighting the transferability of the dimensionality reduction from static equilibrium properties to dynamics.

show abstract

“…To that end, in this work we show how to use as input transient non-equilibrium dynamics in inference of MSMs using Maximum Caliber. Interestingly, using a patchwork of non-equilibrium dynamics is a popular way build MSMs [17][18][19] . Our approach below will be a useful addition to those developed previously.…”

Section: Introductionmentioning

confidence: 99%

Building Markov state models using optimal transport theory

Dixit

Dill

2019

The Journal of Chemical Physics

View full text Add to dashboard Cite

11794Markov State Models (MSMs) describe the rates and routes in conformational dynamics of biomolecules. Computational estimation of MSMs can be expensive because molecular simulations are slow to find and sample the rare transient events.We describe here an efficient approximate way to determine MSM rate matrices by combining Maximum Caliber (maximizing path entropies) with Optimal Transport Theory (minimizing some path cost function, as when routing trucks on transportation networks) to patch together transient dynamical information from multiple nonequilibrium simulations. We give toy examples.

show abstract

Markov state models from short non-equilibrium simulations—Analysis and correction of estimation bias

Cited by 62 publications

References 33 publications

VAMPnets for deep learning of molecular kinetics

VAMPnets for deep learning of molecular kinetics

Interpretable embeddings from molecular simulations using Gaussian mixture variational autoencoders

Building Markov state models using optimal transport theory

Contact Info

Product

Resources

About