Unbiased learning of protein conformational representation via unsupervised random forest DOI Creative Commons

Mohammad Sahil,

Navjeet Ahalawat, Jagannath Mondal

и другие.

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Дек. 1, 2024

Accurate data representation is paramount in biophysics to capture the functionally relevant motions of biomolecules. Traditional feature selection methods, while effective, often rely on labeled based prior knowledge and user-supervision, limiting their applicability novel systems. Here, we present unsupervised random forest (URF), a self-supervised adaptation traditional forests that identifies critical features biomolecules without requiring labels. By devising memory-efficient implementation, first demonstrate URF's capability learn important sets inter-residue protein subsequently resolve its complex conformational landscape, performing at par or surpassing supervised counterpart 15 other leading baseline methods. Crucially, URF supplemented by an internal metric, learning coefficient , which automates process hyper-parameter optimization, making method robust user-friendly. remarkable ability unbiased fashion was validated against 10 independent systems including both folded intrinsically disordered states. In particular, benchmarking investigations showed representations identified are meaningful comparison current state-of-the-art deep As application, show can be seamlessly integrated with downstream analyses pipeline such as Markov state models attain better resolved outputs. The investigation presented here establishes tool for biophysics.

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
Using pretrained graph neural networks with token mixers as geometric featurizers for conformational dynamics DOI

Zihan Pengmei,

Chatipat Lorpaiboon, Spencer C. Guo

и другие.

The Journal of Chemical Physics, Год журнала: 2025, Номер 162(4)

Опубликована: Янв. 28, 2025

Identifying informative low-dimensional features that characterize dynamics in molecular simulations remains a challenge, often requiring extensive manual tuning and system-specific knowledge. Here, we introduce geom2vec, which pretrained graph neural networks (GNNs) are used as universal geometric featurizers. By pretraining equivariant GNNs on large dataset of conformations with self-supervised denoising objective, obtain transferable structural representations useful for learning conformational without further fine-tuning. We show how the learned GNN can capture interpretable relationships between units (tokens) by combining them expressive token mixers. Importantly, decoupling training from downstream tasks enables analysis larger graphs (that represent small proteins at all-atom resolution) limited computational resources. In these ways, geom2vec eliminates need feature selection increases robustness simulation analyses.

Язык: Английский

Процитировано

0
Mapping conformational landscape in protein folding: Benchmarking dimensionality reduction and clustering techniques on the Trp-Cage mini-protein DOI
Sudha Bhattacharya, Suman Chakrabarty

Biophysical Chemistry, Год журнала: 2025, Номер 319, С. 107389 - 107389

Опубликована: Янв. 17, 2025

Язык: Английский

Процитировано

0
Publisher’s Note: “Physically interpretable performance metrics for clustering” [J. Chem. Phys. 161, 244106 (2024)] DOI Open Access

Kinjal Mondal,

Jeffery B. Klauda

The Journal of Chemical Physics, Год журнала: 2025, Номер 162(6)

Опубликована: Фев. 10, 2025

Язык: Английский

Процитировано

0
Computational Analysis of the Fully Activated Orexin Receptor 2 across Various Thermodynamic Ensembles with Surface Tension Monitoring and Markov State Modeling DOI
Rafael Doležal

The Journal of Physical Chemistry B, Год журнала: 2025, Номер unknown

Опубликована: Фев. 11, 2025

In this study, we investigated the stability of fully activated conformation orexin receptor 2 (OX2R) embedded in a pure POPC bilayer using MD simulations. Various thermodynamic ensembles (i.e., NPT, NVT, NVE, NPAT, μVT, and NPγT) were employed to explore dynamical heterogeneity system comprehensive way. addition, informational similarity metrics (e.g., Jensen-Shannon divergence) as well Markov state modeling approaches utilized elucidate kinetics. Special attention was paid assessing surface tension within simulation box, particularly under NPγT conditions, where 21 nominal constants evaluated. Our findings suggest that traditional such NPT may not adequately control physical properties membrane, impacting plausibility OX2R model. general, performed study underscores importance employing ensemble for computational investigations membrane-embedded receptors, it effectively maintains zero simulated system. These results offer valuable insights future research aimed at understanding dynamics designing targeted therapeutics.

Язык: Английский

Процитировано

0
Structural Heterogeneity of Intermediate States Facilitates CRIPT Peptide Binding to the PDZ3 Domain: Insights from Molecular Dynamics and Markov State Model Analysis DOI
Xingyu Song, Dongdong Wang, Jie Ji

и другие.

Journal of Chemical Theory and Computation, Год журнала: 2025, Номер unknown

Опубликована: Фев. 21, 2025

Intrinsically disordered proteins (IDPs), characterized by a lack of defined tertiary structure, are ubiquitous and indispensable components cellular machinery. These participate in diverse array biological processes, often undergoing conformational transitions upon binding to their target, phenomenon termed "folding-upon-binding." The finding raises the question how achieve rapid kinetics presence intrinsic disorder, underlying molecular mechanism remains elusive. This study investigated interaction between C-terminal region CRIPT third PDZ domain PSD-95, critical complex neuronal development. Upon binding, peptide adopts β-strand conformation, process meticulously through extensive dynamics simulations totaling 67.7 μs. Our findings reveal funnel-like landscape which IDPs can adopt multiple conformations prior forming structurally heterogeneous intermediate complexes leading pathways. stabilization these necessitates dynamic interplay native non-native interactions. Markov state model analysis underscores important role structural heterogeneity as it contributes accelerated binding. enrich classical fly-casting provide novel insights into functional advantages conferred disorder.

Язык: Английский

Процитировано

0
Unsupervised learning of structural relaxation in supercooled liquids from short-term fluctuations DOI Creative Commons
Yunrui Qiu, In-Hyuk Jang, Xuhui Huang

и другие.

Proceedings of the National Academy of Sciences, Год журнала: 2025, Номер 122(15)

Опубликована: Апрель 11, 2025

Unraveling the relationship between structural information and dynamic properties of supercooled liquids is one great challenges physics. Dynamic heterogeneity, characterized by propensity particles, often used as a proxy for slowing. Over years, significant efforts have been made to capture variations linked heterogeneity in liquids. In this work, we present an innovative unsupervised machine learning protocol based on time-lagged canonical correlation analysis or autoencoder autonomously identify key order parameter (OP) amorphous structures Kob-Andersen glass former. The OP constructed integrating numerous classical descriptors represents component with strongest short-term timescale thousands times shorter than relaxation time. Strikingly, demonstrates remarkable at long times, significantly outperforming traditional models rivaling supervised models. This that fluctuations contain sufficient about long-time heterogeneity. most important features are density distributions mid-range. As consequence, also exhibits excellent transferability capturing across wide temperature range greatly facilitates evaluation descriptor importance, highlighting its potential broader application other glassy systems.

Язык: Английский

Процитировано

0
Characterization of conformational flexibility in protein structures by applying artificial intelligence to molecular modeling DOI
Kirill Kopylov, Evgeny Kirilin, Vladimir Voevodin

и другие.

Journal of Structural Biology, Год журнала: 2025, Номер unknown, С. 108204 - 108204

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Flow Matching for Optimal Reaction Coordinates of Biomolecular Systems DOI
Mingyuan Zhang, Zhicheng Zhang, Hao Wu

и другие.

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

Опубликована: Дек. 19, 2024

We present flow matching for reaction coordinates (FMRC), a novel deep learning algorithm designed to identify optimal (RC) in biomolecular reversible dynamics. FMRC is based on the mathematical principles of lumpability and decomposability, which we reformulate into conditional probability framework efficient data-driven optimization using generative models. While does not explicitly learn well-established transfer operator or its eigenfunctions, it can effectively encode dynamics leading eigenfunctions system low-dimensional RC space. further quantitatively compare performance with several state-of-the-art algorithms by evaluating quality Markov state models (MSM) constructed their respective spaces, demonstrating superiority three increasingly complex systems. In addition, successfully demonstrated efficacy bias deposition enhanced sampling simple model system. Finally, discuss potential applications downstream such as methods MSM construction.

Язык: Английский

Процитировано

1
Exploration of Intelligent Transformation Path of Traditional Cultural and Creative Products Based on Internet of Things Technology DOI Creative Commons

Jia Xu,

Leilei Zhang

Applied Mathematics and Nonlinear Sciences, Год журнала: 2024, Номер 9(1)

Опубликована: Янв. 1, 2024

Abstract More and more modern technology is applied to traditional cultural creative industries, which enhances the technicality novelty of products, fusion culture profoundly changes people’s way life learning. This paper explores Internet Things (IoT) intelligent transformation process products constructs that integrate voice series. Specifically, a speech recognition model integrated into design designed by analyzing Hidden Markov Model (HMM), proposing solution shortcomings HMM model, choosing perform noise reduction signals through fixed beam algorithm. Based on in this paper, an product with function designed, taking ‘owl wine container’ from Shanxi Museum as example. The regression equation audience’s satisfaction Y (satisfaction) = -0.000263 + 0.208X 1 (practicality) 0.265X 2 (innovation) 0.253X 3 (culture) 0.271X 4 (interactivity) - 0.296X 5 (fun). paper’s have greater impact due their innovativeness interactive function.

Язык: Английский

Процитировано

0