Enhancing subsurface contamination assessment via ensemble prediction of ground electrical property: A Colorado AMD-impacted wetland case study.

Acid mine drainage (AMD) is recognized as a major environmental challenge in the Western United States, particularly in Colorado, leading to extreme subsurface contamination issue. Given Colorado's arid climate and dependence on groundwater, an accurate assessment of AMD-induced contamination is deemed crucial. While in past, machine learning (ML)-based inversion algorithms were used to reconstruct ground electrical properties (GEP) such as relative dielectric permittivity (RDP) from ground penetrating radar (GPR) data for contamination assessment, their inherent non-linear nature can introduce significant uncertainty and non-uniqueness into the reconstructed models. This is a challenge that traditional ML methods are not explicitly designed to address. In this study, a probabilistic hybrid technique has been introduced that combines the DeepLabv3+ architecture-based deep convolutional neural network (DCNN) with an ensemble prediction-based Monte Carlo (MC) dropout method. Different MC dropout rates (1%, 5%, and 10%) were initially evaluated using 1D and 2D synthetic GPR data for accurate and reliable RDP model prediction. The optimal rate was chosen based on minimal prediction uncertainty and the closest alignment of the mean or median model with the true RDP model. Notably, with the optimal MC dropout rate, prediction accuracy of over 95% for the 1D and 2D cases was achieved. Motivated by these results, the hybrid technique was applied to field GPR data collected over an AMD-impacted wetland near Silverton, Colorado. The field results underscored the hybrid technique's ability to predict an accurate subsurface RDP distribution for estimating the spatial extent of AMD-induced contamination. Notably, this technique not only provides a precise assessment of subsurface contamination but also ensures consistent interpretations of subsurface condition by different environmentalists examining the same GPR data. In conclusion, the hybrid technique presents a promising avenue for future environmental studies in regions affected by AMD or other contaminants that alter the natural distribution of GEP.

RNA 3D Structure Prediction: Progress and Perspective.

Ribonucleic acid (RNA) molecules play vital roles in numerous important biological functions such as catalysis and gene regulation. The functions of RNAs are strongly coupled to their structures or proper structure changes, and RNA structure prediction has been paid much attention in the last two decades. Some computational models have been developed to predict RNA three-dimensional (3D) structures in silico, and these models are generally composed of predicting RNA 3D structure ensemble, evaluating near-native RNAs from the structure ensemble, and refining the identified RNAs. In this review, we will make a comprehensive overview of the recent advances in RNA 3D structure modeling, including structure ensemble prediction, evaluation, and refinement. Finally, we will emphasize some insights and perspectives in modeling RNA 3D structures.

Ensemble forecast experiments of summertime sea ice in the Arctic Ocean using the TOPAZ4 ice-ocean data assimilation system.

Precise information on sea ice thickness (SIT) and its prediction at medium-range (2-week) timescale is crucial for the safe maritime navigation in the Arctic Ocean. In this study, we investigate the sensitivity of medium-range prediction skill of summertime SIT distribution in the Arctic marginal seas to atmospheric forecast data, using the 51-member ECMWF operational ensemble prediction system (EPS). For a synoptic-scale cyclone event occurred in July 5-6, 2015, two-week probabilistic forecast experiments were conducted with the TOPAZ4 ice-ocean forecast system, starting on 1st July. The ensemble correlation analysis between the forecast SIT and the meteorological parameters shows that the forecast error of SIT distribution is sensitive to the sea ice drift speed until 1-week, indicating that realistic sea ice drift improves the sea ice thickness prediction. On the other hand, beyond 1 week lead, the forecast error of SIT distribution is more sensitive to surface heat flux rather than sea ice drift. The surface heat flux signal is confined to the sea ice edge region, where the shortwave radiation flux is related to the SIT change through the sea ice melting process. The shortwave radiation flux in the sea ice edge is mostly determined by the sea ice distribution, suggesting that the skillful prediction of sea ice distribution, which is largely affected by synoptic-scale disturbance, at shorter lead times indirectly affects the medium-range forecast skill. A comparison of different ensemble perturbation techniques shows that the prediction skill is better at shorter lead times (up to 1 week), when using an atmospheric EPS rather than the random perturbations used in the operational forecast system, but the random perturbations are advantageous beyond 1 week. Thus, the application of the EPS to an ice-ocean coupled forecast system leads to a more precise sea ice prediction on medium-range timescale, which we expect to become of practical use for the optimum shipping route in the Arctic Ocean.

How to improve infectious disease prediction by integrating environmental data: an application of a novel ensemble analysis strategy to predict HFMD.

This study proposed a novel ensemble analysis strategy to improve hand, foot and mouth disease (HFMD) prediction by integrating environmental data. The approach began by establishing a vector autoregressive model (VAR). Then, a dynamic Bayesian networks (DBN) model was used for variable selection of environmental factors. Finally, a VAR model with constraints (CVAR) was established for predicting the incidence of HFMD in Chengdu city from 2011 to 2017. DBN showed that temperature was related to HFMD at lags 1 and 2. Humidity, wind speed, sunshine, PM10, SO2 and NO2 were related to HFMD at lag 2. Compared with the autoregressive integrated moving average model with external variables (ARIMAX), the CVAR model had a higher coefficient of determination (R2, average difference: + 2.11%; t = 6.2051, P = 0.0003 < 0.05), a lower root mean-squared error (-24.88%; t = -5.2898, P = 0.0007 < 0.05) and a lower mean absolute percentage error (-16.69%; t = -4.3647, P = 0.0024 < 0.05). The accuracy of predicting the time-series shape was 88.16% for the CVAR model and 86.41% for ARIMAX. The CVAR model performed better in terms of variable selection, model interpretation and prediction. Therefore, it could be used by health authorities to identify potential HFMD outbreaks and develop disease control measures.

Identification of interactions of binary variables associated with survival time using survivalFS.

Many medical studies aim to identify factors associated with a time to an event such as survival time or time to relapse. Often, in particular, when binary variables are considered in such studies, interactions of these variables might be the actual relevant factors for predicting, e.g., the time to recurrence of a disease. Testing all possible interactions is often not possible, so that procedures such as logic regression are required that avoid such an exhaustive search. In this article, we present an ensemble method based on logic regression that can cope with the instability of the regression models generated by logic regression. This procedure called survivalFS also provides measures for quantifying the importance of the interactions forming the logic regression models on the time to an event and for the assessment of the individual variables that take the multivariate data structure into account. In this context, we introduce a new performance measure, which is an adaptation of Harrel's concordance index. The performance of survivalFS and the proposed importance measures is evaluated in a simulation study as well as in an application to genotype data from a urinary bladder cancer study. Furthermore, we compare the performance of survivalFS and its importance measures for the individual variables with the variable importance measure used in random survival forests, a popular procedure for the analysis of survival data. These applications show that survivalFS is able to identify interactions associated with time to an event and to outperform random survival forests.

Blind tests of RNA nearest-neighbor energy prediction.

The predictive modeling and design of biologically active RNA molecules requires understanding the energetic balance among their basic components. Rapid developments in computer simulation promise increasingly accurate recovery of RNA's nearest-neighbor (NN) free-energy parameters, but these methods have not been tested in predictive trials or on nonstandard nucleotides. Here, we present, to our knowledge, the first such tests through a RECCES-Rosetta (reweighting of energy-function collection with conformational ensemble sampling in Rosetta) framework that rigorously models conformational entropy, predicts previously unmeasured NN parameters, and estimates these values' systematic uncertainties. RECCES-Rosetta recovers the 10 NN parameters for Watson-Crick stacked base pairs and 32 single-nucleotide dangling-end parameters with unprecedented accuracies: rmsd of 0.28 kcal/mol and 0.41 kcal/mol, respectively. For set-aside test sets, RECCES-Rosetta gives rmsd values of 0.32 kcal/mol on eight stacked pairs involving G-U wobble pairs and 0.99 kcal/mol on seven stacked pairs involving nonstandard isocytidine-isoguanosine pairs. To more rigorously assess RECCES-Rosetta, we carried out four blind predictions for stacked pairs involving 2,6-diaminopurine-U pairs, which achieved 0.64 kcal/mol rmsd accuracy when tested by subsequent experiments. Overall, these results establish that computational methods can now blindly predict energetics of basic RNA motifs, including chemically modified variants, with consistently better than 1 kcal/mol accuracy. Systematic tests indicate that resolving the remaining discrepancies will require energy function improvements beyond simply reweighting component terms, and we propose further blind trials to test such efforts.

Ensemble gene function prediction database reveals genes important for complex I formation in Arabidopsis thaliana.

Recent advances in gene function prediction rely on ensemble approaches that integrate results from multiple inference methods to produce superior predictions. Yet, these developments remain largely unexplored in plants. We have explored and compared two methods to integrate 10 gene co-function networks for Arabidopsis thaliana and demonstrate how the integration of these networks produces more accurate gene function predictions for a larger fraction of genes with unknown function. These predictions were used to identify genes involved in mitochondrial complex I formation, and for five of them, we confirmed the predictions experimentally. The ensemble predictions are provided as a user-friendly online database, EnsembleNet. The methods presented here demonstrate that ensemble gene function prediction is a powerful method to boost prediction performance, whereas the EnsembleNet database provides a cutting-edge community tool to guide experimentalists.

Current and future environmental suitability for bats hosting potential zoonotic pathogens in rural Kenya.

Synanthropic bats live in close proximity to humans and domestic animals, creating opportunities for potential pathogen spillover. We explored environmental correlates of occurrence for a widely distributed synanthropic African bat, Mops pumilus-a species associated with potential zoonotic viruses-and estimated current and future environmental suitability in the Taita Hills region and surrounding plains in Taita-Taveta County in southeast Kenya. To project future environmental suitability, we used four Coupled Model Intercomparison Project Phase 6 general circulation models that capture temperature and precipitation changes for East Africa. The models were parameterized with empirical capture data of M. pumilus collected from 2016 to 2023, combined with satellite-based vegetation, topographic, and climatic data to identify responses to environmental factors. The strongest drivers for current environmental suitability for M. pumilus were short distance to rivers, higher precipitation during the driest months, sparse vegetation-often related to urban areas-and low yearly temperature variation. To predict current and future areas suitable for M. pumilus, we created ensemble niche models, which yielded excellent predictive accuracies. Current suitable environments were located southward from the central and southern Taita Hills and surrounding plains, overlapping with urban centers with the highest human population densities in the area. Future projections for 2050 indicated a moderate increase in suitability range in the southern portion of the region and surrounding plains in human-dominated areas; however, projections for 2090 showed a slight contraction of environmental suitability for M. pumilus, potentially due to the negative impact of increased temperatures. These results show how environmental changes are likely to impact the human exposure risk of bat-borne pathogens and could help public health officials develop strategies to prevent these risks in Taita-Taveta County, Kenya, and other parts of Africa.

FunPredCATH: An ensemble method for predicting protein function using CATH.

MOTIVATION: The growth of unannotated proteins in UniProt increases at a very high rate every year due to more efficient sequencing methods. However, the experimental annotation of proteins is a lengthy and expensive process. Using computational techniques to narrow the search can speed up the process by providing highly specific Gene Ontology (GO) terms. METHODOLOGY: We propose an ensemble approach that combines three generic base predictors that predict Gene Ontology (BP, CC and MF) terms from sequences across different species. We train our models on UniProtGOA annotation data and use the CATH domain resources to identify the protein families. We then calculate a score based on the prevalence of individual GO terms in the functional families that is then used as an indicator of confidence when assigning the GO term to an uncharacterised protein. METHODS: In the ensemble, we use a statistics-based method that scores the occurrence of GO terms in a CATH FunFam against a background set of proteins annotated by the same GO term. We also developed a set-based method that uses Set Intersection and Set Union to score the occurrence of GO terms within the same CATH FunFam. Finally, we also use FunFams-Plus, a predictor method developed by the Orengo Group at UCL to predict GO terms for uncharacterised proteins in the CAFA3 challenge. EVALUATION: We evaluated the methods against the CAFA3 benchmark and DomFun. We used the Precision, Recall and Fmax metrics and the benchmark datasets that are used in CAFA3 to evaluate our models and compare them to the CAFA3 results. Our results show that FunPredCATH compares well with top CAFA methods in the different ontologies and benchmarks. CONTRIBUTIONS: FunPredCATH compares well with other prediction methods on CAFA3, and the ensemble approach outperforms the base methods. We show that non-IEA models obtain higher Fmax scores than the IEA counterparts, while the models including IEA annotations have higher coverage at the expense of a lower Fmax score.

Assessing AF2's ability to predict structural ensembles of proteins.

Recent breakthroughs in protein structure prediction have enhanced the precision and speed at which protein configurations can be determined. Additionally, molecular dynamics (MD) simulations serve as a crucial tool for capturing the conformational space of proteins, providing valuable insights into their structural fluctuations. However, the scope of MD simulations is often limited by the accessible timescales and the computational resources available, posing challenges to comprehensively exploring protein behaviors. Recently emerging approaches have focused on expanding the capability of AlphaFold2 (AF2) to predict conformational substates of protein. Here, we benchmark the performance of various workflows that have adapted AF2 for ensemble prediction and compare the obtained structures with ensembles obtained from MD simulations and NMR. We provide an overview of the levels of performance and accessible timescales that can currently be achieved with machine learning (ML) based ensemble generation. Significant minima of the free energy surfaces remain undetected.

A Novel Deep Learning-Based Classification Framework for COVID-19 Assisted with Weighted Average Ensemble Modeling.

COVID-19 is an infectious disease caused by the deadly virus SARS-CoV-2 that affects the lung of the patient. Different symptoms, including fever, muscle pain and respiratory syndrome, can be identified in COVID-19-affected patients. The disease needs to be diagnosed in a timely manner, otherwise the lung infection can turn into a severe form and the patient's life may be in danger. In this work, an ensemble deep learning-based technique is proposed for COVID-19 detection that can classify the disease with high accuracy, efficiency, and reliability. A weighted average ensemble (WAE) prediction was performed by combining three CNN models, namely Xception, VGG19 and ResNet50V2, where 97.25% and 94.10% accuracy was achieved for binary and multiclass classification, respectively. To accurately detect the disease, different test methods have been proposed and developed, some of which are even being used in real-time situations. RT-PCR is one of the most successful COVID-19 detection methods, and is being used worldwide with high accuracy and sensitivity. However, complexity and time-consuming manual processes are limitations of this method. To make the detection process automated, researchers across the world have started to use deep learning to detect COVID-19 applied on medical imaging. Although most of the existing systems offer high accuracy, different limitations, including high variance, overfitting and generalization errors, can be found that can degrade the system performance. Some of the reasons behind those limitations are a lack of reliable data resources, missing preprocessing techniques, a lack of proper model selection, etc., which eventually create reliability issues. Reliability is an important factor for any healthcare system. Here, transfer learning with better preprocessing techniques applied on two benchmark datasets makes the work more reliable. The weighted average ensemble technique with hyperparameter tuning ensures better accuracy than using a randomly selected single CNN model.

Multi-source dynamic ensemble prediction of infectious disease and application in COVID-19 case.

Background: The development of an epidemic always exhibits multiwave oscillation owing to various anthropogenic sources of transmission. Particularly in populated areas, the large-scaled human mobility led to the transmission of the virus faster and more complex. The accurate prediction of the spread of infectious diseases remains a problem. To solve this problem, we propose a new method called the multi-source dynamic ensemble prediction (MDEP) method that incorporates a modified susceptible-exposed-infected-removed (SEIR) model to improve the accuracy of the prediction result. Methods: The modified SEIR model is based on the compartment model, which is suitable for local-scale and confined spaces, where human mobility on a large scale is not considered. Moreover, compartmental models cannot be used to predict multiwave epidemics. The proposed MDEP method can remedy defects in the compartment model. In this study, multi-source prediction was made on the development of coronavirus disease 2019 (COVID-19) and dynamically assembled to obtain the final integrated result. We used the real epidemic data of COVID-19 in three cities in China: Beijing, Lanzhou, and Beihai. Epidemiological data were collected from 17 April, 2022 to 12 August, 2022. Results: Compared to the one-wave modified SEIR model, the MDEP method can depict the multiwave development of COVID-19. The MDEP method was applied to predict the number of cumulative cases of recent COVID-19 outbreaks in the aforementioned cities in China. The average accuracy rates in Beijing, Lanzhou, and Beihai were 89.15%, 91.74%, and 94.97%, respectively. Conclusions: The MDEP method improved the prediction accuracy of COVID-19. With further application to other infectious diseases, the MDEP method will provide accurate predictions of infectious diseases and aid governments make appropriate directives.

Predicting habitat suitability for Ixodes ricinus and Ixodes persulcatus ticks in Finland.

BACKGROUND: Ticks are responsible for transmitting several notable pathogens worldwide. Finland lies in a zone where two human-biting tick species co-occur: Ixodes ricinus and Ixodes persulcatus. Tick densities have increased in boreal regions worldwide during past decades, and tick-borne pathogens have been identified as one of the major threats to public health in the face of climate change. METHODS: We used species distribution modelling techniques to predict the distributions of I. ricinus and I. persulcatus, using aggregated historical data from 2014 to 2020 and new tick occurrence data from 2021. By aiming to fill the gaps in tick occurrence data, we created a new sampling strategy across Finland. We also screened for tick-borne encephalitis virus (TBEV) and Borrelia from the newly collected ticks. Climate, land use and vegetation data, and population densities of the tick hosts were used in various combinations on four data sets to estimate tick species' distributions across mainland Finland with a 1-km resolution. RESULTS: In the 2021 survey, 89 new locations were sampled of which 25 new presences and 63 absences were found for I. ricinus and one new presence and 88 absences for I. persulcatus. A total of 502 ticks were collected and analysed; no ticks were positive for TBEV, while 56 (47%) of the 120 pools, including adult, nymph, and larva pools, were positive for Borrelia (minimum infection rate 11.2%, respectively). Our prediction results demonstrate that two combined predictor data sets based on ensemble mean models yielded the highest predictive accuracy for both I. ricinus (AUC = 0.91, 0.94) and I. persulcatus (AUC = 0.93, 0.96). The suitable habitats for I. ricinus were determined by higher relative humidity, air temperature, precipitation sum, and middle-infrared reflectance levels and higher densities of white-tailed deer, European hare, and red fox. For I. persulcatus, locations with greater precipitation and air temperature and higher white-tailed deer, roe deer, and mountain hare densities were associated with higher occurrence probabilities. Suitable habitats for I. ricinus ranged from southern Finland up to Central Ostrobothnia and North Karelia, excluding areas in Ostrobothnia and Pirkanmaa. For I. persulcatus, suitable areas were located along the western coast from Ostrobothnia to southern Lapland, in North Karelia, North Savo, Kainuu, and areas in Pirkanmaa and Päijät-Häme. CONCLUSIONS: This is the first study conducted in Finland that estimates potential tick species distributions using environmental and host data. Our results can be utilized in vector control strategies, as supporting material in recommendations issued by public health authorities, and as predictor data for modelling the risk for tick-borne diseases.

Tracking glioblastoma progression after initial resection with minimal reaction-diffusion models.

We describe a preliminary effort to model the growth and progression of glioblastoma multiforme, an aggressive form of primary brain cancer, in patients undergoing treatment for recurrence of tumor following initial surgery and chemoradiation. Two reaction-diffusion models are used: the Fisher-Kolmogorov equation and a 2-population model, developed by the authors, that divides the tumor into actively proliferating and quiescent (or necrotic) cells. The models are simulated on 3-dimensional brain geometries derived from magnetic resonance imaging (MRI) scans provided by the Barrow Neurological Institute. The study consists of 17 clinical time intervals across 10 patients that have been followed in detail, each of whom shows significant progression of tumor over a period of 1 to 3 months on sequential follow up scans. A Taguchi sampling design is implemented to estimate the variability of the predicted tumors to using 144 different choices of model parameters. In 9 cases, model parameters can be identified such that the simulated tumor, using both models, contains at least 40 percent of the volume of the observed tumor. We discuss some potential improvements that can be made to the parameterizations of the models and their initialization.

Evaluating the added value of multi-input atmospheric transport ensemble modeling for applications of the Comprehensive Nuclear Test-Ban Treaty organization (CTBTO).

The Comprehensive Nuclear Test-Ban Treaty Organization (CTBTO) runs to date operationally an atmospheric transport modeling chain in backward mode based on operational deterministic European Centre for Medium-Range Weather Forecasts-Integrated Forecasting System (ECMWF-IFS) and on National Centers for Environmental Prediction-Global Forecast System (NCEP-GFS) input data. Meanwhile, ensemble dispersion modeling is becoming more and more widespread due to the ever increasing computational power and storage capacities. The potential benefit of this approach for current and possible future CTBTO applications was investigated using data from the ECMWF-Ensemble Prediction System (EPS). Five different test cases - among which are the ETEX-I experiment and the Fukushima accident - were run in backward or forward mode and - in the light of a future operational application - special emphasis was put on the performance of an arbitrarily selected 10- versus the full 51-member ensemble. For those test cases run in backward mode and based on a puff release it became evident that Possible Source Regions (PSRs) can be meaningfully reduced in size compared to results based solely on the deterministic run by applying minimum and probability of exceedance ensemble metrics. It was further demonstrated that a given puff release of 4E10 Bq of Se-75 can be reproduced within the meteorological uncertainty range [1.9E9 Bq,1.7E13 Bq] including a probability for not exceeding an assumed upper limit source term using simple scaling of a measurement with the corresponding ensemble metrics of backward fields. For the test cases run in forward mode it was found that the control run as well as 10- and 51-member medians all exhibit similar performance in time series evaluation. Maximum rank difference adds up to less than 10% with reference to possible rank values [0,4]. The maximum difference in the Brier score for both ensembles is less than 3%. The main added value of the ensemble lies in producing meteorologically induced concentration uncertainties and thus explaining observed measurements at specific sites. Depending on the specific test case and on the ensemble size between 27 and 74% of samples all lie within concentration ranges derived from the different meteorological fields used. In the future uncertainty information per sample could be used in a full source term inversion to account for the meteorological uncertainty in a proper way. It can be concluded that a 10-member meteorological ensemble is good enough to already benefit from useful ensemble properties. Meteorological uncertainty to a large degree is covered by the 10-member subset because forecast uncertainty is largely suppressed due to concatenating analyses and short term forecasts, as required in the operational CTBTO procedure, on which this study focuses. Besides, members from different analyses times are on average unrelated. It was recommended to Working Group B of CTBTO to implement the ensemble system software in the near future.

Adaptive ensemble perception tracking.

Recently, tracking models based on bounding box regression (such as region proposal networks), built on the Siamese network, have attracted much attention. Despite their promising performance, these trackers are less effective in perceiving the target information in the following two aspects. First, existing regression models cannot take a global view of a large-scale target since the effective receptive field of a neuron is too small to cover the target with a large scale. Second, the neurons with a fixed receptive field (RF) size in these models cannot adapt to the scale and aspect ratio changes of the target. In this paper, we propose an adaptive ensemble perception tracking framework to address these issues. Specifically, we first construct a per-pixel prediction model, which predicts the target state at each pixel of the correlated feature. On top of the per-pixel prediction model, we then develop a confidence-guided ensemble prediction mechanism. The ensemble mechanism adaptively fuses the predictions of multiple pixels with the guidance of confidence maps, which enlarges the perception range and enhances the adaptive perception ability at the object-level. In addition, we introduce a receptive field adaption model to enhance the adaptive perception ability at the neuron-level, which adjusts the RF by adaptively integrating the features with different RFs. Extensive experimental results on the VOT2018, VOT2016, UAV123, LaSOT, and TC128 datasets demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods in terms of accuracy and speed.

Uncertainties in atmospheric dispersion modelling during nuclear accidents.

Predictions of the atmospheric dispersion of radionuclides accidentally released from a nuclear power plant are influenced by two large sources of uncertainty: one associated with the meteorological data employed, and one with the source term, i.e. the temporal evolution of the amount and physical and chemical properties of the release. A methodology is presented for quantitative estimation of the variability of the prediction of atmospheric dispersion resulting from both sources of uncertainty. The methodology, which allows for efficient calculation, and thus is well suited for real-time assessment, is applied to a hypothetical accidental release of radionuclides.

Forecasting of Landslide Displacement Using a Probability-Scheme Combination Ensemble Prediction Technique.

Data-driven models have been extensively employed in landslide displacement prediction. However, predictive uncertainty, which consists of input uncertainty, parameter uncertainty, and model uncertainty, is usually disregarded in deterministic data-driven modeling, and point estimates are separately presented. In this study, a probability-scheme combination ensemble prediction that employs quantile regression neural networks and kernel density estimation (QRNNs-KDE) is proposed for robust and accurate prediction and uncertainty quantification of landslide displacement. In the ensemble model, QRNNs serve as base learning algorithms to generate multiple base learners. Final ensemble prediction is obtained by integration of all base learners through a probability combination scheme based on KDE. The Fanjiaping landslide in the Three Gorges Reservoir area (TGRA) was selected as a case study to explore the performance of the ensemble prediction. Based on long-term (2006-2018) and near real-time monitoring data, a comprehensive analysis of the deformation characteristics was conducted for fully understanding the triggering factors. The experimental results indicate that the QRNNs-KDE approach can perform predictions with perfect performance and outperform the traditional backpropagation (BP), radial basis function (RBF), extreme learning machine (ELM), support vector machine (SVM) methods, bootstrap-extreme learning machine-artificial neural network (bootstrap-ELM-ANN), and Copula-kernel-based support vector machine quantile regression (Copula-KSVMQR). The proposed QRNNs-KDE approach has significant potential in medium-term to long-term horizon forecasting and quantification of uncertainty.

MCP: A multi-component learning machine to predict protein secondary structure.

The Gene or DNA sequence in every cell does not control genetic properties on its own; Rather, this is done through the translation of DNA into protein and subsequent formation of a certain 3D structure. The biological function of a protein is tightly connected to its specific 3D structure. Prediction of the protein secondary structure is a crucial intermediate step towards elucidating its 3D structure and function. Traditional experimental methods for prediction of protein structure are expensive and time-consuming. Nevertheless, the average accuracy of the suggested solutions has hardly reached beyond 80%. The possible underlying reasons are the ambiguous sequence-structure relation, noise in input protein data, class imbalance, and the high dimensionality of the encoding schemes. Furthermore, we utilize a compound string dissimilarity measure to directly interpret protein sequence content and avoid information loss. In order to improve accuracy, we employ two different classifiers including support vector machine and fuzzy nearest neighbor and collectively aggregate the classification outcomes to infer the final protein structures. We conduct comprehensive experiments to compare our model with the current state-of-the-art approaches. The experimental results demonstrate that given a set of input sequences, our multi-component framework can accurately predict the protein structure. Nevertheless, the effectiveness of our unified model can be further enhanced through framework configuration.

Assessment of freshwater discharge into a coastal bay through multi-basin ensemble hydrological modelling.

Coastal basins of the Brittany peninsula (France) are hydrological hot spots. A high level of nutrient pollution affects many of these basins and causes algal blooms in several coastal bays; nine have been a specific focus of the European Commission since 2007. The flux of each contributing basin flowing into these bays must be examined to assess the conditions and explore mitigation options. However, this task encounters a large lack of data since most of the basins are ungauged. In this context, this study developed a method which facilitates transposition of hydrographs from gauged basins to ungauged neighbouring basins of interest. Inverting a simple geomorphology-based transfer function of the gauged basin which describes travel time through channels enables the net rainfall time-series to be estimated from the discharge time-series of donor basins. To estimate the net rainfall of a given ungauged catchment, several net rainfall time series of gauged catchments are averaged. The resulting net rainfall is then transposed onto the ungauged target basin and convoluted by its own transfer function to estimate the hydrograph. This allows the transposition of as many hydrographs as there are different donor basins. This ensemble prediction enables the proportion of prediction uncertainty that is due to the heterogeneity in hydrological behaviour to be estimated. Moreover, the time-series of the donor basins are combined to estimate the ungauged net rainfall time-series. This provides a discharge prediction which values all available measurements. The method was applied to the highly controversial Saint Brieuc Bay, where it was possible to quantify the contribution of each coastal basin, even those influenced by dams, and ultimately the entire volume of fresh water entering the bay at an hourly time step. This work opens perspectives to additionally refine estimation of the associated nutrient fluxes.