Variational Selection of Features for Molecular Kinetics

Abstract

The modeling of atomistic biomolecular simulations using kinetic models such as Markov state models (MSMs) has had many notable algorithmic advances in recent years. The variational principle has opened the door for a nearly fully automated toolkit for selecting models that predict the long-time kinetics from molecular dynamics simulations. However, one yet-unoptimized step of the pipeline involves choosing the features, or collective variables, from which the model should be constructed. In order to build intuitive models, these collective variables are often sought to be interpretable and familiar features, such as torsional angles or contact distances in a protein structure. However, previous approaches for evaluating the chosen features rely on constructing a full MSM, which in turn requires additional hyperparameters to be chosen, and hence leads to a computationally expensive framework. Here, we present a method to optimize the feature choice directly, without requiring the construction of the final kinetic model. We demonstrate our rigorous preprocessing algorithm on a canonical set of twelve fast-folding protein simulations, and show that our procedure leads to more efficient model selection.

13 pages, 8 figures

References in corpus (6)

• Identification of slow molecular order parameters for Markov model construction
• Variational cross-validation of slow dynamical modes in molecular kinetics
• Variational Koopman models: slow collective variables and molecular kinetics from short off-equilibrium simulations
• Variational approach for learning Markov processes from time series data
• Identification of kinetic order parameters for non-equilibrium dynamics
• Spectral rate theory for projected two-state kinetics

Cited by in corpus (17)

• Machine learning for molecular simulation
• TorchMD: A deep learning framework for molecular simulations
• The Exploration of Chemical Reaction Networks
• Coarse Graining Molecular Dynamics with Graph Neural Networks
• Special Topic: Markov Models of Molecular Kinetics
• GraphVAMPNet, using graph neural networks and variational approach to markov processes for dynamical modeling of biomolecules
• Correlation-based feature selection to identify functional dynamics in proteins
• A Deep Autoencoder Framework for Discovery of Metastable Ensembles in Biomacromolecules
• Recent advances in describing and driving crystal nucleation using machine learning and artificial intelligence
• Selecting Features for Markov Modeling: A Case Study on HP35
• Towards a Benchmark for Markov State Models: The Folding of HP35
• Deeptime: a Python library for machine learning dynamical models from time series data
• Deflation reveals dynamical structure in nondominant reaction coordinates
• Identifying the leading dynamics of ubiquitin: a comparison between the tICA and the LE4PD slow fluctuations in amino acids' position
• Volume-Scaled Common Nearest Neighbor Clustering Algorithm with Free-Energy Hierarchy
• Improving Estimation of the Koopman Operator with Kolmogorov-Smirnov Indicator Functions
• Contact cluster modeling of allosteric communication in PDZ domains