Physically interpretable performance metrics for clustering DOI

Kinjal Mondal,

Jeffery B. Klauda

The Journal of Chemical Physics, Год журнала: 2024, Номер 161(24)

Опубликована: Дек. 26, 2024

Clustering is a type of machine learning technique, which used to group huge amounts data based on their similarity into separate groups or clusters. very important task that nowadays analyze the and diverse amount coming out molecular dynamics (MD) simulations. Typically, from MD simulations in terms various frames trajectory are clustered different representative element each studied separately. Now, question this process is: what quality clusters obtained? There several performance metrics available literature such as silhouette index Davies–Bouldin Index often clustering. However, most these focus overlap reduced dimension for clustering do not physically properties parameters system. To address issue, we have developed two interpretable scoring physical system analyzing. We tested our algorithm three systems: (1) Ising model, (2) peptide folding unfolding WT HP35, (3) protein–ligand an enzyme substrate, (4) dissociated trajectory. show provide us match with intuition about systems.

Язык: Английский

Augmenting Human Expertise in Weighted Ensemble Simulations through Deep Learning-Based Information Bottleneck DOI
Dedi Wang, Pratyush Tiwary

Journal of Chemical Theory and Computation, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 26, 2024

The weighted ensemble (WE) method stands out as a widely used segment-based sampling technique renowned for its rigorous treatment of kinetics. WE framework typically involves initially mapping the configuration space onto low-dimensional collective variable (CV) and then partitioning it into bins. efficacy simulations heavily depends on selection CVs binning schemes. recently proposed State Predictive Information Bottleneck (SPIB) has emerged promising tool automatically constructing from data guiding enhanced through an iterative manner. In this work, we advance data-driven pipeline by incorporating prior expert knowledge. Our hybrid approach combines SPIB-learned to enhance in explored regions with expert-based guide exploration interest, synergizing strengths both methods. Through benchmarking alanine dipeptide chignoin systems, demonstrate that our effectively guides sample states reduces run-to-run variances. Moreover, integration SPIB model also enhances analysis interpretation simulation identifying metastable pathways, offering direct visualization dynamics.

Язык: Английский

Процитировано

4
Graph Neural Network-State Predictive Information Bottleneck (GNN-SPIB) approach for learning molecular thermodynamics and kinetics DOI Creative Commons

Ziyue Zou,

Dedi Wang, Pratyush Tiwary

и другие.

Digital Discovery, Год журнала: 2024, Номер 4(1), С. 211 - 221

Опубликована: Ноя. 28, 2024

We present a graph-based differentiable representation learning method from atomic coordinates for enhanced sampling methods to learn both thermodynamic and kinetic properties of system.

Язык: Английский

Процитировано

2
Advances and Challenges in Milestoning Simulations for Drug–Target Kinetics DOI Creative Commons
Anupam Anand Ojha, Lane Votapka, Rommie E. Amaro

и другие.

Journal of Chemical Theory and Computation, Год журнала: 2024, Номер 20(22), С. 9759 - 9769

Опубликована: Ноя. 7, 2024

Molecular dynamics simulations have become indispensable for exploring complex biological processes, yet their limitations in capturing rare events hinder our understanding of drug–target kinetics. In this Perspective, we investigate the domain milestoning to understand challenge. The approach divides phase space into discrete cells, offering extended time scale insights. This Perspective traces history, applications, and future potential context It explores fundamental principles milestoning, highlighting importance probabilistic transitions transition independence. Markovian with Voronoi tessellations is revisited address traditional challenges. While observing advancements field, also addresses impending challenges estimating unbinding rate constants through simulations, paving way more effective drug design strategies.

Язык: Английский

Процитировано

1
Analysis of transition rates from variational flooding using analytical theory DOI

David Cummins,

Carter Longstreth,

James McCarty

и другие.

The Journal of Chemical Physics, Год журнала: 2024, Номер 161(19)

Опубликована: Ноя. 18, 2024

Variational flooding is an enhanced sampling method for obtaining kinetic rates from molecular dynamics simulations. This inspired by the idea of conformational that employs a boost potential acting along chosen reaction coordinate to accelerate rare events. In this work, we show how empirical distribution crossing times variational simulations can be modeled with analytical Kramers' time-dependent rate (KTR) theory. An optimized bias fills metastable free energy basins constructed variationally (VES) method. VES-derived then augmented switching function determines fill level boost. Having prescribed gives expression KTR theory used extract unbiased rates. case static potential, barrier follows expected exponential distribution, and are extracted series boosted at discrete levels. Introducing increases gradually over simulation time leads simplified procedure fitting biased We demonstrate approach paradigmatic cases alanine dipeptide in vacuum, asymmetric SN2 reaction, folding chignolin explicit solvent.

Язык: Английский

Процитировано

0
Therapeutic Potential and Mechanistic Insights of a Novel Synthetic α-Lactalbumin-Derived Peptide for the Treatment of Liver Fibrosis DOI
Sara Maher, Shimaa Atta,

Manal Kamel

и другие.

Journal of Clinical and Experimental Hepatology, Год журнала: 2024, Номер 15(3), С. 102488 - 102488

Опубликована: Дек. 15, 2024

Язык: Английский

Процитировано

0
Physically interpretable performance metrics for clustering DOI

Kinjal Mondal,

Jeffery B. Klauda

The Journal of Chemical Physics, Год журнала: 2024, Номер 161(24)

Опубликована: Дек. 26, 2024

Clustering is a type of machine learning technique, which used to group huge amounts data based on their similarity into separate groups or clusters. very important task that nowadays analyze the and diverse amount coming out molecular dynamics (MD) simulations. Typically, from MD simulations in terms various frames trajectory are clustered different representative element each studied separately. Now, question this process is: what quality clusters obtained? There several performance metrics available literature such as silhouette index Davies–Bouldin Index often clustering. However, most these focus overlap reduced dimension for clustering do not physically properties parameters system. To address issue, we have developed two interpretable scoring physical system analyzing. We tested our algorithm three systems: (1) Ising model, (2) peptide folding unfolding WT HP35, (3) protein–ligand an enzyme substrate, (4) dissociated trajectory. show provide us match with intuition about systems.

Язык: Английский

Процитировано

0