Stochastic Neural Network Approach for Learning High-Dimensional Free Energy Surfaces

Schneider, Elia; Dai, Luke; Topper, Robert Q.; Drechsel-Grau, Christof; Tuckerman, Mark E.

doi:10.1103/physrevlett.119.150601

Cited by 106 publications

(114 citation statements)

References 37 publications

Supporting

Mentioning

111

Contrasting

Unclassified

Order By: Relevance

“…The problem of dimensionality reduction in complex phase-space transitions has recently seen an increasing number of approaches based on statistical projection of data, 17,18 as well as methods that involve machine learning/artificial intelligence approaches. 19 For drug design applications, the first kind of approach appears to be more useful: it keeps the role of every observable employed in the dimensionality reduction in a transparent way, allowing a straightforward interpretation of the final results.…”

Section: Introductionmentioning

confidence: 99%

Chasing the full free energy landscape of neuroreceptor/ligand unbinding by metadynamics simulations

Capelli

Bochicchio

Piccini

et al. 2019

Preprint

View full text Add to dashboard Cite

Predicting the complete free energy landscape associated with protein-ligand unbinding would greatly help designing drugs with highly optimized pharmacokinetics.Here we investigate the unbinding of the iperoxo agonist to its target human neuroreceptor M 2 , embedded in a neuronal membrane. By feeding out-of-equilibrium molecular simulations data in a classification analysis, we identify the few essential reaction coordinates of the process. The full landscape is then reconstructed using an exact enhanced sampling method, well-tempered metadynamics in its funnel variant. The calculations reproduce well the measured affinity, provide a rationale for mutagenesis data and show that the ligand can escape via two different routes. The allosteric modulator LY2119620 turns out to hamper both escapes routes, thus slowing down the unbinding process, as experimentally observed. This computationally affordable protocol is totally general and it can be easily applied to determine the full free energy landscape of membrane receptors/drug interactions.

show abstract

Section: Introductionmentioning

confidence: 99%

Chasing the full free energy landscape of neuroreceptor/ligand unbinding by metadynamics simulations

Capelli

Bochicchio

Piccini

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…In the last few years, the deep learning, a particular variant of machine learning approach, is emerging as an e cient approach for identification of linear and/or nonlinear reaction coordinates to perform biased sampling [57][58][59][60] . In future, it would be interesting to compare the behavior (i.e.…”

Section: Discussionmentioning

confidence: 99%

Assessment and Optimization of Collective Variables for Protein Conformational Landscape: GB1 β-hairpin as a Case Study

Ahalawat

Mondal

2018

Preprint

View full text Add to dashboard Cite

Collective variables (CV), when chosen judiciously, can play an important role in recognizing rate-limiting processes and rare events in any biomolecular systems. However, high dimensionality and inherent complexities associated with such biochemical systems render the identification of an optimal CV a challenging task, which in turn precludes the elucidation of underlying conformational landscape in su cient details. In this context, a relevant model system is presented by 16-residue -hairpin of GB1 protein. Despite being the target of numerous theoretical and computational studies for understanding the protein folding, the set of CVs optimally characterizing the conformational landscape of -hairpin of GB1 protein has remained elusive, resulting in a lack of consensus on its folding mechanism. Here we address this by proposing a pair of optimal CVs which can resolve the underlying free energy landscape of GB1 hairpin quite e ciently. Expressed as a linear combination of a number of traditional CVs, the optimal CV for this system is derived by employing recently introduced Time-structured Independent Component Analysis (TICA) approach on a large number of independent unbiased simulations. By projecting the replica-exchange simulated trajectories along these pair of optimized CVs, the resulting free energy landscape of this system are able to resolve four distinct well-separated metastable states encompassing the extensive ensembles of folded,unfolded and molten globule states. Importantly, the optimized CVs were found to be capable of automatically recovering a novel partial helical state of this protein, without needing to explicitly invoke helicity as a constituent CV. Furthermore, a quantitative sensitivity analysis of each constituent in the optimized CV provided key insights on the relative contributions of the constituent CVs in the overall free energy landscapes. Finally, the kinetic pathways connecting these metastable states, constructed using a Markov State Model, provide an optimum description of underlying folding mechanism of the peptide. Taken together, this work o↵ers a quantitatively robust approach towards comprehensive mapping of the underlying folding landscape of a quintessential model system along its optimized collective variables.

show abstract

“…Efficient exploration of configuration space by adding an adaptively computed biasing potential using machine learning to the original dynamics. [31]- [35] 16. Use of the "information Bottleneck" approach to design an ANN that will identify a collective coordinate that will guide simulations with importance sampling to correct bias [36], [37].This leads to a collective coordinator with good physical (chemical) interpretation.…”

Section: A Mlautotuninghpc -Smart Ensemblesmentioning

confidence: 99%

Learning Everywhere: A Taxonomy for the Integration of Machine Learning and Simulations

Fox

Jha

2019

2019 15th International Conference on eScience (eScience)

View full text Add to dashboard Cite

We present a taxonomy of research on Machine Learning (ML) applied to enhance simulations together with a catalog of some activities. We cover eight patterns for the link of ML to the simulations or systems plus three algorithmic areas: particle dynamics, agent-based models and partial differential equations. The patterns are further divided into three action areas: Improving simulation with Configurations and Integration of Data, Learn Structure, Theory and Model for Simulation, and Learn to make Surrogates.

show abstract

Stochastic Neural Network Approach for Learning High-Dimensional Free Energy Surfaces

Cited by 106 publications

References 37 publications

Chasing the full free energy landscape of neuroreceptor/ligand unbinding by metadynamics simulations

Chasing the full free energy landscape of neuroreceptor/ligand unbinding by metadynamics simulations

Assessment and Optimization of Collective Variables for Protein Conformational Landscape: GB1 β-hairpin as a Case Study

Learning Everywhere: A Taxonomy for the Integration of Machine Learning and Simulations

Contact Info

Product

Resources

About