Weighted Ensemble Research Articles

Insights into Translocation of Arginine-Rich Cell-Penetrating Peptides across a Model Membrane.

It is well-known that membrane deformation and water pores contribute to the spontaneous translocation of arginine-rich cell-penetrating peptides (CPPs). We confirm this through the observation of the spontaneous translocation of single R9 (nona-arginine) and Tat (48-60) peptides across a model membrane using the weighted ensemble (WE) method within all-atom molecular dynamics (MD) simulations. Furthermore, we demonstrate that membrane deformation and the presence of a water pore reduce the effective charge of the CPP and the bending rigidity of the model membrane during translocation. We find that R9 disturbs the model membrane more than Tat (48-60), leading to more efficient translocation of R9 across the model membrane.

Rare-Event Sampling using a Reinforcement Learning-Based Weighted Ensemble Method.

Despite the power of path sampling strategies in enabling simulations of rare events, such strategies have not reached their full potential. A common challenge that remains is the identification of a progress coordinate that captures the slow relevant motions of a rare event. Here we have developed a weighted ensemble (WE) path sampling strategy that exploits reinforcement learning to automatically identify an effective progress coordinate among a set of potential coordinates during a simulation. We apply our WE strategy with reinforcement learning to three benchmark systems: (i) an egg carton-shaped toy potential, (ii) an S-shaped toy potential, and (iii) a dimer of the HIV-1 capsid protein (C-terminal domain). To enable rapid testing of the latter system at the atomic level, we employed discrete-state synthetic molecular dynamics trajectories using a generative, fine-grained Markov state model that was based on extensive conventional simulations. Our results demonstrate that using concepts from reinforcement learning with a weighted ensemble of trajectories automatically identifies relevant progress coordinates among multiple candidates at a given time during a simulation. Due to the rigorous weighting of trajectories, the simulations maintain rigorous kinetics.

Ice Coating Prediction Based on Two-Stage Adaptive Weighted Ensemble Learning

Severe ice accretion on transmission lines can disrupt electrical grids and compromise the stability of power systems. Consequently, precise prediction of ice coating on transmission lines is vital for guiding their operation and maintenance. Traditional single-model icing prediction methods often exhibit limited accuracy under varying environmental conditions and fail to yield highly accurate predictions. We propose a multi-scenario, two-stage adaptive ensemble strategy (MTAES) for ice coating prediction to address this issue. A combined clustering approach is employed to refine the division of ice weather scenarios, segmenting historical samples into multiple scenarios. Within each scenario, the bagging approach generates multiple training subsets, with the extreme learning machine (ELM) used to build diverse models. Subsequently, a two-stage adaptive weight allocation mechanism is introduced. This mechanism calculates the distance from the scenario cluster centers and the prediction error of similar samples in the validation set for each test sample. Weights are dynamically allocated based on these data, leading to the final output results through an adaptive ensemble from the base model repository. The experimental results show that the model is significantly better than traditional models in predicting ice thickness. Key indicators of RMSE, MAE, and R2 reach 0.675, 0.522, and 83.2%, respectively, verifying the effectiveness of multi-scene partitioning and adaptive weighting methods in improving the accuracy of ice cover prediction.

BAD-NEUS: Rapidly converging trajectory stratification.

An issue for molecular dynamics simulations is that events of interest often involve timescales that are much longer than the simulation time step, which is set by the fastest timescales of the model. Because of this timescale separation, direct simulation of many events is prohibitively computationally costly. This issue can be overcome by aggregating information from many relatively short simulations that sample segments of trajectories involving events of interest. This is the strategy of Markov state models (MSMs) and related approaches, but such methods suffer from approximation error because the variables defining the states generally do not capture the dynamics fully. By contrast, once converged, the weighted ensemble (WE) method aggregates information from trajectory segments so as to yield unbiased estimates of both thermodynamic and kinetic statistics. Unfortunately, errors decay no faster than unbiased simulation in WE as originally formulated and commonly deployed. Here, we introduce a theoretical framework for describing WE that shows that the introduction of an approximate stationary distribution on top of the stratification, as in nonequilibrium umbrella sampling (NEUS), accelerates convergence. Building on ideas from MSMs and related methods, we generalize the NEUS approach in such a way that the approximation error can be reduced systematically. We show that the improved algorithm can decrease the simulation time required to achieve the desired precision by orders of magnitude.

WEDAP: A Python Package for Streamlined Plotting of Molecular Simulation Data.

Given the growing interest in path sampling methods for extending the time scales of molecular dynamics (MD) simulations, there has been great interest in software tools that streamline the generation of plots for monitoring the progress of large-scale simulations. Here, we present the WEDAP Python package for simplifying the analysis of data generated from either conventional MD simulations or the weighted ensemble (WE) path sampling method, as implemented in the widely used WESTPA software package. WEDAP facilitates (i) the parsing of WE simulation data stored in highly compressed, hierarchical HDF5 files and (ii) incorporates trajectory weights from WE simulations into all generated plots. Our Python package consists of multiple user-friendly interfaces: a command-line interface, a graphical user interface, and a Python application programming interface. We demonstrate the plotting features of WEDAP through a series of examples using data from WE and conventional MD simulations that focus on the HIV-1 capsid protein's C-terminal domain dimer as a showcase system. The source code for WEDAP is freely available on GitHub at https://github.com/chonglab-pitt/wedap.

Development of maximum relevant prior feature ensemble (MRPFE) index to characterize future drought using global climate models

Drought is one of the foremost outcomes of global warming and global climate change. It is a serious threat to humans and other living beings. To reduce the adverse impact of drought, mitigation strategies as well as sound projections of extreme events are essential. This research aims to strengthen the robustness of anticipated twenty-first century drought by combining different Global Climate Models (GCMs). In this article, we develop a new drought index, named Maximum Relevant Prior Feature Ensemble index that is based on the newly proposed weighting scheme, called weighted ensemble (WE). In the application, this study considers 32 randomly scattered grid points within the Tibetan Plateau region and 18 GCMs of Coupled Model Intercomparison Project Phase 6 (CMIP6) of precipitation. In this study, the comparative inferences of the WE scheme are made with the traditional simple model averaging (SMA). To investigate the trend and long-term probability of various classes, this research employs Markov chain steady states probability, Mann–Kendall trend test, and Sen’s Slope estimator. The outcomes of this research are twofold. Firstly, the comparative inference shows that the proposed weighting scheme has greater efficiency than SMA to conflate GCMs. Secondly, the research indicates that the Tibetan Plateau is projected to experience “moderate drought (MD)” in the twenty-first century.

Ensemble distributional forecasting for insurance loss reserving

ABSTRACT Loss reserving generally focuses on identifying a single model that can generate superior predictive performance. However, different loss reserving models specialise in capturing different aspects of loss data. This is recognised in practice in the sense that results from different models are often considered, and sometimes combined. For instance, actuaries may take a weighted average of the prediction outcomes from various loss reserving models, often based on subjective assessments. In this paper, we propose a systematic framework to objectively combine (i.e. ensemble) multiple stochastic loss reserving models such that the strengths offered by different models can be utilised effectively. Our framework contains two main innovations compared to existing literature and practice. Firstly, our criteria model combination considers the full distributional properties of the ensemble and not just the central estimate – which is of particular importance in the reserving context. Secondly, our framework is that it is tailored for the features inherent to reserving data. These include, for instance, accident, development, calendar, and claim maturity effects. Crucially, the relative importance and scarcity of data across accident periods renders the problem distinct from the traditional ensembling techniques in statistical learning. Our framework is illustrated with a complex synthetic dataset. In the results, the optimised ensemble outperforms both (i) traditional model selection strategies, and (ii) an equally weighted ensemble. In particular, the improvement occurs not only with central estimates but also relevant quantiles, such as the 75th percentile of reserves (typically of interest to both insurers and regulators). The framework developed in this paper can be implemented thanks to an R package, ADLP, which is available from CRAN.

Mechanism of Daunomycin Intercalation into DNA from Enhanced Sampling Simulations.

Daunomycin is a widely used anticancer drug, yet the mechanism underlying how it binds to DNA remains contested. 469 all-atom trajectories of daunomycin binding to the DNA oligonucleotide d(GCG CAC GTG CGC) were collected using weighted ensemble (WE)-enhanced sampling. Mechanistic insights were revealed through analysis of the ensemble of trajectories. Initially, the binding process involves a ubiquitous hydrogen bond between the DNA backbone and the NH3+ group on daunomycin. During the binding process, most trajectories exhibited similar structural changes to DNA, including DNA base pair rise, bending, and minor groove width changes. Variability within the ensemble of binding trajectories illuminates differences in the orientation of daunomycin as it initially intercalates; around 10% of trajectories needed minimal rearrangement from intercalation to reaching the fully bound configuration, whereas most needed an additional 1-5 ns to rearrange. The results here emphasize the utility of generating an ensemble of trajectories to discern biomolecular binding mechanisms.

Revisiting Textbook Azide-Clock Reactions: A "Propeller-Crawling" Mechanism Explains Differences in Rates.

An ongoing challenge to chemists is the analysis of pathways and kinetics for chemical reactions in solution, including transient structures between the reactants and products that are difficult to resolve using laboratory experiments. Here, we enabled direct molecular dynamics simulations of a textbook series of chemical reactions on the hundreds of ns to μs time scale using the weighted ensemble (WE) path sampling strategy with hybrid quantum mechanical/molecular mechanical (QM/MM) models. We focused on azide-clock reactions involving addition of an azide anion to each of three long-lived trityl cations in an acetonitrile-water solvent mixture. Results reveal a two-step mechanism: (1) diffusional collision of reactants to form an ion-pair intermediate; (2) "activation" or rearrangement of the intermediate to the product. Our simulations yield not only reaction rates that are within error of experiment but also rates for individual steps, indicating the activation step as rate-limiting for all three cations. Further, the trend in reaction rates is due to dynamical effects, i.e., differing extents of the azide anion "crawling" along the cation's phenyl-ring "propellers" during the activation step. Our study demonstrates the power of analyzing pathways and kinetics to gain insights on reaction mechanisms, underscoring the value of including WE and other related path sampling strategies in the modern toolbox for chemists.

Bibimbap : Pre-trained models ensemble for Domain Generalization

This paper addresses a machine learning problem often challenged by differences in the distributions of training and real-world data. We propose a framework that addresses the problem of underfitting in the ensembling method using pre-trained models and improves the performance and robustness of deep learning models through ensemble diversity. For the naive weight ensembling framework, we discovered that the ensembled models could not lie in the same loss basin under extreme domain shift conditions, suggesting that a loss barrier may exist. We used a fine-tuning step after the weighted ensemble to address the underfitting problem caused by the loss barrier and stabilize the batch normalization running parameters. We also inferred through qualitative analysis that the diversity of ensemble models affects domain generalization. We validate our method on a large-scale image dataset (ImageNet-1K) and chemical molecule data, which is suitable for testing with domain shift problems due to its data-splitting method.

Fuzzy Divergence Weighted Ensemble Clustering With Spectral Learning Based on Random Projections for Big Data

Fuzzy Divergence Weighted Ensemble Clustering With Spectral Learning Based on Random Projections for Big Data

Ensemble Region-Specific GMMs for Subduction Earthquakes

Abstract This study develops data-driven global and region-specific ground-motion models (GMMs) for subduction earthquakes using a weighted average ensemble model to combine four different nonparametric supervised machine-learning (ML) algorithms, including an artificial neural network, a kernel ridge regressor, a random forest regressor, and a support vector regressor. To achieve this goal, we train individual models using a subset of the Next Generation Attenuation-Subduction (NGA-Sub) data set, including 9559 recordings out of 153 interface and intraslab earthquakes recorded at 3202 different stations. A grid search is used to find each model’s best hyperparameters. Then, we use an equally weighted average ensemble approach to combine these four models. Ensemble modeling is a technique that combines the strengths of multiple ML algorithms to mitigate their weaknesses. The ensemble model considers moment magnitude (M), rupture distance (Rrup), time-averaged shear-wave velocity in the upper 30 m (VS30), and depth to the top of the rupture plane (Ztor), as well as tectonic and region as input parameters, and predicts various median orientation-independent horizontal component ground-motion intensity measures such as peak ground displacement, peak ground velocity, peak ground acceleration, and 5%-damped pseudospectral acceleration values at spectral periods of 0.01–10 s in log scale. Although no functional form is defined, the response spectra and the distance and magnitude scaling trends of the weighted average ensemble model are consistent and comparable with the NGA-Sub GMMs, with slightly lower standard deviations. A mixed effects regression analysis is used to partition the total aleatory variability into between-event, between-station, and event-site-corrected components. The derived global GMMs are applicable to interface earthquakes with M 4.9–9.12, 14≤Rrup≤1000 km, and Ztor≤47 km for sites having VS30values between 95 and 2230 m/s. For intraslab events, the derived global GMMs are applicable to M 4.0–8.0, 28≤Rrup≤1000 km, and 30≤Ztor≤200 km for sites having VS30 values between 95 and 2100 m/s.

Integrating Electron Paramagnetic Resonance Spectroscopy and Computational Modeling to Measure Protein Structure and Dynamics.

Electron paramagnetic resonance (EPR) has become a powerful probe of conformational heterogeneity and dynamics of biomolecules. In this Review, we discuss different computational modeling techniques that enrich the interpretation of EPR measurements of dynamics or distance restraints. A variety of spin labels are surveyed to provide a background for the discussion of modeling tools. Molecular dynamics (MD) simulations of models containing spin labels provide dynamical properties of biomolecules and their labels. These simulations can be used to predict EPR spectra, sample stable conformations and sample rotameric preferences of label sidechains. For molecular motions longer than milliseconds, enhanced sampling strategies and de novo prediction software incorporating or validated by EPR measurements are able to efficiently refine or predict protein conformations, respectively. To sample large-amplitude conformational transition, a coarse-grained or an atomistic weighted ensemble (WE) strategy can be guided with EPR insights. Looking forward, we anticipate an integrative strategy for efficient sampling of alternate conformations by de novo predictions, followed by validations by systematic EPR measurements and MD simulations. Continuous pathways between alternate states can be further sampled by WE-MD including all intermediate states.

Direct observation of negative cooperativity in a detoxification enzyme at the atomic level by Electron Paramagnetic Resonance spectroscopy and simulation.

The catalytic activity of human glutathione S-transferase A1-1 (hGSTA1-1), a homodimeric detoxification enzyme, is dependent on the conformational dynamics of a key C-terminal helix α9 in each monomer. However, the structural details of how the two monomers interact upon binding of substrates is not well understood and the structure of the ligand-free state of the hGSTA1-1 homodimer has not been resolved. Here, we used a combination of electron paramagnetic resonance (EPR)distance measurements and weighted ensemble (WE) simulations to characterize the conformational ensemble of the ligand-free state at the atomic level. EPR measurements reveal a broad distance distribution between a pair of Cu(II) labels in the ligand-free state that gradually shifts and narrows as a function of increasing ligand concentration. These shifts suggest changes in the relative positioning of the two α9 helices upon ligand binding. WE simulations generated unbiased pathways for the seconds-timescale transition between alternate states of the enzyme, leading to the generation of atomically detailed structures of the ligand-free state. Notably, the simulations provide direct observations of negative cooperativity between the monomers of hGSTA1-1, which involve the mutually exclusive docking of α9 in each monomer as a lid over the active site. We identify key interactions between residues that lead to this negative cooperativity. Negative cooperativity may be essential for interaction of hGSTA1-1 with a wide variety of toxic substrates and their subsequent neutralization. More broadly, this work demonstrates the power of integrating EPR distances with WE rare-events sampling strategy to gain mechanistic information on protein function at the atomic level.

Dynamic Resilience in Credit Card Fraud Detection: The Adaptive Accuracy Weighted Ensemble Approach

This paper introduces the Adaptive Accuracy Weighted Ensemble (AAWE) algorithm, strategically devised to fortify credit card fraud detection against the constant threat of concept drift. Centered on the foundational Multilayer Perceptron (MLP) classifier, AAWE seamlessly incorporates dynamic weighting mechanisms to skillfully acclimate to the ever-shifting landscape of fraudulent activities. This innovative progression substantiates AAWE's capacity to substantially augment the precision of credit card fraud detection, demonstrating adeptness in negotiating the intricate challenges presented by perpetually evolving fraudulent strategies. In the realm of credit card fraud detection, wherein subtle deviations in fraud methodologies can yield profound consequences, AAWE's intrinsic adaptability emerges as a pivotal asset. The algorithm's ability to promptly respond to concept drift nuances becomes a critical edge. With empirical evidence showcasing its commendable accuracy rates, AAWE emerges as a compelling remedy to the exigent concept drift quandary within credit card fraud detection. This research sheds a compelling light on AAWE's potential to deliver robust and pinpoint-accurate fraud detection outcomes, solidifying its stature as an invaluable instrument for real-world scenarios where timely accommodation to emerging fraud patterns is of paramount significance.

Worth the Weight: Sub-Pocket EXplorer (SubPEx), a Weighted Ensemble Method to Enhance Binding-Pocket Conformational Sampling.

Structure-based virtual screening (VS) is an effective method for identifying potential small-molecule ligands, but traditional VS approaches consider only a single binding-pocket conformation. Consequently, they struggle to identify ligands that bind to alternate conformations. Ensemble docking helps address this issue by incorporating multiple conformations into the docking process, but it depends on methods that can thoroughly explore pocket flexibility. We here introduce Sub-Pocket EXplorer (SubPEx), an approach that uses weighted ensemble (WE) path sampling to accelerate binding-pocket sampling. As proof of principle, we apply SubPEx to three proteins relevant to drug discovery: heat shock protein 90, influenza neuraminidase, and yeast hexokinase 2. SubPEx is available free of charge without registration under the terms of the open-source MIT license: http://durrantlab.com/subpex/.

Robustness of gridded precipitation products for vietnam basins using the comprehensive assessment framework of rainfall

The use of satellite–based precipitation products (SPPs) have become increasingly prevalent as key inputs to provide regional rainfall for improving hydrological simulations in data–spare regions. This study introduces a new approach – Comprehensive Assessment Framework of Rainfall (CAFR) to evaluate six satellite–based precipitation products (SPPs) for eleven basins with different sizes across Vietnam (2007–2015). These SPPs include the Global Precipitation Mission (GPM) Integrated Multi-satellitE Retrievals for Global Precipitation Measurement Final run Version 6 (GPM IMERGF V6), Multi–Source Weighted–Ensemble Precipitation (MSWEP) V2.2, Soil Moisture to Rain (SM2RAIN) – Advanced SCATterometer (ASCAT) V1.5, Asian Precipitation–Highly–Resolved Observational Data Integration Towards Evaluation (APHRODITE) V1901, Climate Hazards group Infrared Precipitation with Stations (CHIRPS) V2.0, and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) – Climate Data Record (CDR) V1.0. With the proposed CAFR: (1) IMERGF is suggested to have the best performance overall, especially when simulating flood peaks; (2) SM2RAIN–ASCAT demonstrates the best skills in metrics related to the dry season; (3) For streamflow simulations, SPPs' performance is sensitive to basin size, with larger basins showing better performance skills. In this study, we demonstrate the capability of our proposed framework to better understand SPP applications in hydrological modeling.

Shapley Values as a Strategy for Ensemble Weights Estimation

This study introduces a novel performance-based weighting scheme for ensemble learning using the Shapley value. The weighting uses the reciprocal of binary cross-entropy as a base learner’s performance metric and estimates its Shapley value to measure the overall contribution of a learner to an equally weighted ensemble of various sizes. Two variants of this strategy were empirically compared with a single monolith model and other static weighting strategies using two large banking-related datasets. A variant that discards learners with a negative Shapley value was ranked as first or at least second when constructing homogeneous ensembles, whereas for heterogeneous ensembles this strategy resulted in a better or at least similar detection performance to other weighting strategies tested. The main limitation being the computational complexity of Shapley calculations, the explored weighting strategy could be considered as a generalization of performance-based weighting.

Predictive Churn Modeling for Sustainable Business in the Telecommunication Industry: Optimized Weighted Ensemble Machine Learning

Customers are prominent resources in every business for its sustainability. Therefore, predicting customer churn is significant for reducing churn, particularly in the high-churn-rate telecommunications business. To identify customers at risk of churning, tactical marketing actions can be strategized to raise the likelihood of the churn-probable customers remaining as customers. This might provide a corporation with significant savings. Hence, in this work, a churn prediction system is developed to assist telecommunication operators in detecting potential churn customers. In the proposed framework, the input data quality is improved through the processes of exploratory data analysis and data preprocessing for identifying data errors and comprehending data patterns. Then, feature engineering and data sampling processes are performed to transform the captured data into an appropriate form for classification and imbalanced data handling. An optimized ensemble learning model is proposed for classification in this framework. Unlike other ensemble models, the proposed classification model is an optimized weighted soft voting ensemble with a sequence of weights applied to weigh the prediction of each base learner with the hypothesis that specific base learners in the ensemble have more skill than others. In this optimization, Powell’s optimization algorithm is applied to optimize the ensemble weights of influence according to the base learners’ importance. The efficiency of the proposed optimally weighted ensemble learning model is evaluated in a real-world database. The empirical results show that the proposed customer churn prediction system achieves a promising performance with an accuracy score of 84% and an F1 score of 83.42%. Existing customer churn prediction systems are studied. We achieved a higher prediction accuracy than the other systems, including machine learning and deep learning models.

Molecular Mechanism of Ligand Binding to the Minor Groove of DNA

Although DNA–ligand binding is pervasive in biology, little is known about molecular-level binding mechanisms. Using all-atom, explicit-solvent molecular dynamics simulations in conjunction with weighted ensemble (WE)-enhanced sampling, an ensemble of 2562 binding trajectories of Hoechst 33258 (H33258) to d(CGC AAA TTT GCG) was generated from which the binding mechanism was extracted. In particular, the electrostatic interaction between the positively charged H33258 and the negatively charged DNA backbone drives the formation of initial H33258–DNA contacts. After this initial contact, a hinge-like intermediate state is formed in which one end of H33258 inserts into the minor groove of DNA. Following hinge state formation is a concerted motion whereby the second end of H33258 swings into the minor groove and the spine of hydration along the minor groove causing dehydration. This study illustrates how WE-enhanced simulations of biomolecular ligation processes can offer novel mechanistic insights by generating ensembles of binding events.