WeTICA: A directed search weighted ensemble based enhanced sampling method to estimate rare event kinetics in a reduced dimensional space

The Journal of Chemical Physics, Год журнала: 2025, Номер 162(3)

Опубликована: Янв. 15, 2025

Estimating rare event kinetics from molecular dynamics simulations is a non-trivial task despite the great advances in enhanced sampling methods. Weighted Ensemble (WE) simulation, special class of techniques, offers way to directly calculate kinetic rate constants biased trajectories without need modify underlying energy landscape using bias potentials. Conventional WE algorithms use different binning schemes partition collective variable (CV) space separating two metastable states interest. In this work, we have developed new "binless" simulation algorithm bypass hurdles optimizing procedures. Our proposed protocol (WeTICA) uses low-dimensional CV drive toward specified target state. We applied recover unfolding three proteins: (A) TC5b Trp-cage mutant, (B) TC10b and (C) Protein G, with times spanning range between 3 40 μs projections along predefined fixed Time-lagged Independent Component Analysis (TICA) eigenvectors as CVs. Calculated converge reported values good accuracy more than one order magnitude less cumulative time scales or priori knowledge CVs that can capture unfolding. be used other linear CVs, not limited TICA. Moreover, walker selection criteria for resampling employed on sophisticated nonlinear further improvements binless

Язык: Английский

---

Exploring the Conformational Ensembles of Protein–Protein Complex with Transformer-Based Generative Model

Journal of Chemical Theory and Computation, Год журнала: 2024, Номер 20(11), С. 4469 - 4480

Опубликована: Май 30, 2024

Protein–protein interactions are the basis of many protein functions, and understanding contact conformational changes protein–protein is crucial for linking structure to biological function. Although difficult detect experimentally, molecular dynamics (MD) simulations widely used study ensembles complexes, but there significant limitations in sampling efficiency computational costs. In this study, a generative neural network was trained on complex conformations obtained from directly generate novel with physical realism. We demonstrated use deep learning model based transformer architecture explore complexes through MD simulations. The results showed that learned latent space can be unsampled obtaining new complementing pre-existing ones, which as an exploratory tool analysis enhancement complexes.

Язык: Английский

---

Integration of a Randomized Sequence Scanning Approach in AlphaFold2 and Local Frustration Profiling of Conformational States Enable Interpretable Atomistic Characterization of Conformational Ensembles and Detection of Hidden Allosteric States in the ABL1 Protein Kinase

Journal of Chemical Theory and Computation, Год журнала: 2024, Номер 20(12), С. 5317 - 5336

Опубликована: Июнь 12, 2024

Despite the success of AlphaFold methods in predicting single protein structures, these showed intrinsic limitations characterization multiple functional conformations allosteric proteins. The recent NMR-based structural determination unbound ABL kinase active state and discovery inactive low-populated that are unique for present an ideal challenge AlphaFold2 approaches. In current study, we employ several adaptations methodology to predict conformational ensembles states including randomized alanine sequence scanning combined with alignment subsampling proposed this study. We show new adaptation local frustration profiling enables accurate prediction structures ensembles, also offering a robust approach interpretable predictions detection hidden states. found large high residue clusters uniquely characteristic low-populated, fully form can define energetically frustrated cracking sites transitions, presenting difficult targets AlphaFold2. results study uncovered previously unappreciated fundamental connections between profiles ability This integration landscape-based analysis allows atomistic providing physical basis successes detecting play significant role regulation.

Язык: Английский

---

Multi-Objective Design of DNA-Stabilized Nanoclusters Using Variational Autoencoders With Automatic Feature Extraction

ACS Nano, Год журнала: 2024, Номер unknown

Опубликована: Сен. 17, 2024

DNA-stabilized silver nanoclusters (Ag

Язык: Английский

---

Reading the repertoire: Progress in adaptive immune receptor analysis using machine learning

Cell Systems, Год журнала: 2024, Номер 15(12), С. 1168 - 1189

Опубликована: Дек. 1, 2024

Язык: Английский

---

Deep learning guided design of dynamic proteins

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

Опубликована: Июль 19, 2024

Deep learning has greatly advanced design of highly stable static protein structures, but the controlled conformational dynamics that are hallmarks natural switch-like signaling proteins have remained inaccessible to

Язык: Английский

---

The evolution of the Amber additive protein force field: History, current status, and future

The Journal of Chemical Physics, Год журнала: 2025, Номер 162(3)

Опубликована: Янв. 16, 2025

Molecular dynamics simulations are pivotal in elucidating the intricate properties of biological molecules. Nonetheless, reliability their outcomes hinges on precision molecular force field utilized. In this perspective, we present a comprehensive review developmental trajectory Amber additive protein field, delving into researchers’ persistent quest for higher fields and prevailing challenges. We detail parameterization process fields, emphasizing specific improvements retained features each version compared to predecessors. Furthermore, discuss challenges that current encounter balancing interactions protein–protein, protein–water, water–water simulations, as well potential solutions overcome these issues.

Язык: Английский

---

Frontiers in integrative structural modeling of macromolecular assemblies

QRB Discovery, Год журнала: 2025, Номер 6

Опубликована: Янв. 1, 2025

Abstract Integrative modeling enables structure determination for large macromolecular assemblies by combining data from multiple experiments with theoretical and computational predictions. Recent advancements in AI-based prediction cryo electron-microscopy have sparked renewed enthusiasm integrative modeling; structures methods can be integrated situ maps to characterize assemblies. This approach previously allowed us others determine the architectures of diverse assemblies, such as nuclear pore complexes, chromatin remodelers, cell–cell junctions. Experimental spanning several scales was used these studies, ranging high-resolution data, X-ray crystallography AlphaFold structure, low-resolution cryo-electron tomography co-immunoprecipitation experiments. Two recurrent challenges emerged across a range studies. First, contained significant fractions disordered regions, necessitating development new regions context ordered regions. Second, needed developed utilize information tomography, timely challenge structural biology is increasingly moving towards characterization. Here, we recapitulate recent developments proteins analysis highlight other opportunities method modeling.

Язык: Английский

---

A Deep Learning Method for Predicting Interactions for Intrinsically Disordered Regions of Proteins

Опубликована: Янв. 1, 2025

Язык: Английский

---

Use of AI-methods over MD simulations in the sampling of conformational ensembles in IDPs

Frontiers in Molecular Biosciences, Год журнала: 2025, Номер 12

Опубликована: Апрель 8, 2025

Intrinsically Disordered Proteins (IDPs) challenge traditional structure-function paradigms by existing as dynamic ensembles rather than stable tertiary structures. Capturing these is critical to understanding their biological roles, yet Molecular Dynamics (MD) simulations, though accurate and widely used, are computationally expensive struggle sample rare, transient states. Artificial intelligence (AI) offers a transformative alternative, with deep learning (DL) enabling efficient scalable conformational sampling. They leverage large-scale datasets learn complex, non-linear, sequence-to-structure relationships, allowing for the modeling of in IDPs without constraints physics-based approaches. Such DL approaches have been shown outperform MD generating diverse comparable accuracy. Most models rely primarily on simulated data training experimental serves role validation, aligning generated observable physical biochemical properties. However, challenges remain, including dependence quality, limited interpretability, scalability larger proteins. Hybrid combining AI can bridge gaps integrating statistical thermodynamic feasibility. Future directions include incorporating observables into frameworks refine predictions enhance applicability. AI-driven methods hold significant promise IDP research, offering novel insights protein dynamics therapeutic targeting while overcoming limitations simulations.

Язык: Английский

---