Applicability Domain Of Model Research Articles

Neural-based approaches to overcome feature selection and applicability domain in drug-related property prediction

In the fields of pharmaceutical research and biomedical sciences, QSAR modeling is an established approach during drug discovery for prediction of biological activity of drug candidates. Yet, QSAR modeling poses a series of open challenges. First, chemical compounds are represented on a high-dimensional space and thus feature selection is typically applied, although this task entails a challenging combinatorial problem with potential loss of information. Second, the definition of the applicability domain of a QSAR model is a desirable aspect to determine the reliability of predictions on unseen chemicals, which is often difficult to assess due to the extent of the chemical space. Finally, interpretability of these models is also a critical issue for drug designers. The purpose of this work is to thoroughly assess the application of neural-based methods and recent advances deep learning for QSAR modeling. We hypothesize that neural-based methods can overcome the need to perform a descriptor selection phase. We developed three QSAR models based on neural networks for prediction of relevant chemical and biomedical properties that, in the absence of any feature selection step, can outperform the state-of-the-art models for such properties. We also implemented an embedded applicability domain technique based on network output probabilities that proved to be effective; its application improved the predictive performance of the model. Finally, we proposed the use of a post hoc feature analysis technique based on an aggregation of network weights, which enabled effective detection of relevant features in the model.

ProtoSteer: Steering Deep Sequence Model with Prototypes.

Recently we have witnessed growing adoption of deep sequence models (e.g. LSTMs) in many application domains, including predictive health care, natural language processing, and log analysis. However, the intricate working mechanism of these models confines their accessibility to the domain experts. Their black-box nature also makes it a challenging task to incorporate domain-specific knowledge of the experts into the model. In ProtoSteer (Prototype Steering), we tackle the challenge of directly involving the domain experts to steer a deep sequence model without relying on model developers as intermediaries. Our approach originates in case-based reasoning, which imitates the common human problem-solving process of consulting past experiences to solve new problems. We utilize ProSeNet (Prototype Sequence Network), which learns a small set of exemplar cases (i.e., prototypes) from historical data. In ProtoSteer they serve both as an efficient visual summary of the original data and explanations of model decisions. With ProtoSteer the domain experts can inspect, critique, and revise the prototypes interactively. The system then incorporates user-specified prototypes and incrementally updates the model. We conduct extensive case studies and expert interviews in application domains including sentiment analysis on texts and predictive diagnostics based on vehicle fault logs. The results demonstrate that involvements of domain users can help obtain more interpretable models with concise prototypes while retaining similar accuracy.

An in vitro and in silico investigation of human pregnane X receptor agonistic activity of poly- and perfluorinated compounds using the heuristic method–best subset and comparative similarity indices analysis

Poly- and perfluorinated compounds (PFCs) may induce potential endocrine-disrupting hormonal effects. However, the molecular mechanism of the toxicology of PFCs remains unclear, and the insufficient information is available on the biological activities of PFCs at present. In this study, the cell-based reporter gene assays were used to determine the agonistic activity of PFCs on the human pregnane X receptor (hPXR). The heuristic method combined with best subset modeling (HM-BSM) based on Dragon descriptors and comparative similarity indices analysis (CoMSIA) were employed to build classical quantitative structure-activity relationship (QSAR) and three-dimensional QSAR models, respectively. The applicability domain (AD) of the classical QSAR model was assessed. Both the HM-BSM and CoMSIA approaches demonstrated good robustness, predictive ability, and mechanistic interpretability. The r2 and leave-one-out cross-validation squared correlated coefficient (q2LOO) values were 0.872 and 0.759 for the HM-BSM, and 0.976 and 0.751 for the CoMSIA model, respectively. The hPXR agonistic activity of the PFCs predicted by the built HM-BSM and CoMSIA agreed well with experimental activity, with root mean square error (RMSE) values of 0.0803 and 0.117, respectively, and external validation squared correlated coefficients (q2EXT) of 0.972 and 0.932, respectively. The hPXR agonistic activity of PFCs was related to their molecular polarizability, charge and atomic mass. Hydrogen bonding and hydrophobic interactions constituted the primary intermolecular forces between PFCs and the hPXR. The developed models were used to screen the PFCs with high hPXR agonistic activity.

Ecotoxicological QSAR modelling of organic chemicals against Pseudokirchneriella subcapitata using consensus predictions approach

ABSTRACTThe present study provides robust consensus quantitative structure-activity relationship (QSAR) models developed from 334 organic chemicals covering a wide chemical domain for the prediction of effective concentrations of chemicals for 50% and 10% inhibition of algal growth. Only 2D descriptors with definite physicochemical meaning were employed for QSAR model building, whereas development, validation and interpretation were achieved following the strict Organization for Economic Co-operation and Development (OECD) recommended guidelines. Genetic algorithm along with stepwise approach was used in feature selection while the final QSAR models were derived using partial least squares regression technique. The applicability domain of the developed models was also checked. The obtained consensus models were then used to predict 64 organic chemicals having no definite observed responses while the confidence of predictions was checked by the ‘prediction reliability indicator’ tool. The developed models should be applicable for data gap filling in case of new or untested organic chemicals provided they fall within the domain of the model and can also be implemented to design safer alternatives to the environment.

Modeling oil-brine interfacial tension at high pressure and high salinity conditions

Accurate estimation of interfacial tension (IFT) in crude oil/brine system is of great importance for many processes in petroleum and chemical engineering. The current study plays emphasis on introducing the “Gradient Boosting Decision Tree (GBDT)” and ‘‘Adaptive Boosting Support Vector Regression (AdaBoost SVR)” as novel powerful machine learning tools to determine the IFT of crude oil/brine system. Two sorts of models have been developed using each of these two data-driven methods. The first kind includes six inputs, namely pressure (P), temperature (T) and four parameters describing the proprieties of crude oil (total acid number (TAN) and specific gravity (SG) and brine (NaCl equivalent salinity (Seq) and pH), while the second kind deals with four inputs (without including pH and TAN). To this end, an extensive databank including 560 experimental points was considered, in which 80% of the points were employed for the training phase and the remaining part was utilized as blind test data. Results revealed that the proposed approaches provide very satisfactory predictions, and the implemented GBDT model with six inputs is the most accurate model of all with an average absolute relative error of 1.01%. Moreover, the outcomes of the GBDT model are better than literature models. Finally, outlier diagnostic using Leverage approach was performed to investigate the applicability domain of the GBDT model and to evaluate the quality of employed data.

Support Vector Machine model for hERG inhibitory activities based on the integrated hERG database using descriptor selection by NSGA-II

Assessing the hERG liability in the early stages of drug discovery programs is important. The recent increase of hERG-related information in public databases enabled various successful applications of machine learning techniques to predict hERG inhibition. However, most of these researches constructed the datasets from only one database, limiting the predictability and scope of the models. In this study, a hERG classification model was constructed using the largest dataset for hERG inhibition built by integrating multiple databases. The integrated dataset consisted of more than 291,000 structurally diverse compounds derived from ChEMBL, GOSTAR, PubChem, and hERGCentral. The prediction model was built by support vector machine (SVM) with descriptor selection based on Non-dominated Sorting Genetic Algorithm-II (NSGA-II) to optimize the descriptor set for maximum prediction performance with the minimal number of descriptors. The SVM classification model using 72 selected descriptors and ECFP_4 structural fingerprints recorded kappa statistics of 0.733 and accuracy of 0.984 for the test set, substantially outperforming the prediction performance of the current commercial applications for hERG prediction. Finally, the applicability domain of the prediction model was assessed based on the molecular similarity between the training set and test set compounds.

FAME 3: Predicting the Sites of Metabolism in Synthetic Compounds and Natural Products for Phase 1 and Phase 2 Metabolic Enzymes.

In this work we present the third generation of FAst MEtabolizer (FAME 3), a collection of extra trees classifiers for the prediction of sites of metabolism (SoMs) in small molecules such as drugs, druglike compounds, natural products, agrochemicals, and cosmetics. FAME 3 was derived from the MetaQSAR database ( Pedretti et al. J. Med. Chem. 2018 , 61 , 1019 ), a recently published data resource on xenobiotic metabolism that contains more than 2100 substrates annotated with more than 6300 experimentally confirmed SoMs related to redox reactions, hydrolysis and other nonredox reactions, and conjugation reactions. In tests with holdout data, FAME 3 models reached competitive performance, with Matthews correlation coefficients (MCCs) ranging from 0.50 for a global model covering phase 1 and phase 2 metabolism, to 0.75 for a focused model for phase 2 metabolism. A model focused on cytochrome P450 metabolism yielded an MCC of 0.57. Results from case studies with several synthetic compounds, natural products, and natural product derivatives demonstrate the agreement between model predictions and literature data even for molecules with structural patterns clearly distinct from those present in the training data. The applicability domains of the individual models were estimated by a new, atom-based distance measure (FAMEscore) that is based on a nearest-neighbor search in the space of atom environments. FAME 3 is available via a public web service at https://nerdd.zbh.uni-hamburg.de/ and as a self-contained Java software package, free for academic and noncommercial research.

Estimation of predictive performance for test data in applicability domains using y‐randomization

AbstractA new measure of predictive performance for objective variables in regression analysis is proposed, enabling the y‐errors of new samples or test samples to be estimated in the applicability domains (ADs) of regression models. The proposed measure, based on y‐randomization, considers chance correlations and is calculated using only training data. This chance correlation‐excluded mean absolute error (MAECCE) can estimate the y‐errors of new samples considering the influence of chance correlations in the given dataset on the regression models. Experiments using numerical simulation, quantitative structure‐activity relationship, and quantitative structure‐property relationship datasets confirm that MAECCE can estimate the distribution of y‐errors of new samples in ADs for various training datasets, descriptor sets, and regression analysis methods, enabling chance correlations to be eliminated from data analysis. Python and MATLAB codes for the proposed algorithm are available at https://github.com/hkaneko1985/maecce.

Improvement of predictive accuracy in semi-supervised regression analysis by selecting unlabeled chemical structures

While regression models can predict the values of objective variables y, such as properties and activities, from the values of explanatory variables X, such as molecular descriptors and experimental conditions, the reliability of the predicted y-values depends on the X-values of the samples being similar to those of the training samples. When the number of training samples is low, the diversity of the training data is also low, which leads to narrow applicability domains (ADs) of regression models. To enlarge these ADs, models are constructed using both labeled and unlabeled data, a process known as semi-supervised learning. However, if outliers and missing values are included in the unlabeled data, the prediction performance of the models will decrease. Therefore, we propose a method for selecting unlabeled data that lie within the ADs. In addition, we only use unlabeled samples for which there is a high probability that the predicted y-values exist in the target ranges of y in the ADs. The probability is calculated by Bayesian optimization based on the means and variances of the y-values estimated from a Gaussian process. By analyzing a quantitative structure–activity relationship dataset and a quantitative structure–property relationship dataset, informative unlabeled samples can be selected, allowing the estimation performance of regression models to be improved by using the proposed methods.

Development of Predictive Antioxidant Models for 1,3,4-Oxadiazoles by Quantitative Structure Activity Relationship

The free radical scavenging properties of 1,3,4-oxadiazoles have been explored by the application of quantitative structure activity relationship (QSAR) studies. The entire data set of the oxadiazole derivatives were minimized and subsequently optimized at the density functional theory (DFT) level in combination with the Becke's three-parameter Lee-Yang-Parr hybrid functional (B3LYP) hybrid functional and 6-311G* basis set. Kennard Stone algorithm was employed in data division into training and test sets. The training set were employed in QSAR model development by genetic function algorithm (GFA), while the test set were used to validate the developed models. The applicability domain of the developed model was accessed by the leverage approach. The varation inflation factor, degree of contribution and mean effect of each descriptor were calculated. Quantum chemical and molecular descriptors were generated for each molecule in the data set. Five predictive models that met all the requirements for acceptability with good validation results were developed. The best of the five models gave the following validation results: , , and c , rmsep . The QSAR analysis revealed that the sum of e-state descriptors of strength for potential hydrogen bonds of path length 9 (SHBint9) and topological radius (topoRadius) are the most crucial descriptors that influence the free radical scavenging activities of 1,3,4-oxadiazole derivatives.

Molecular docking and receptor-based QASR studies on pyrimidine derivatives as potential phosphodiesterase 10A inhibitors

In the present work, molecular docking methodology in combination with quantitative structure–activity relationship (QSAR) was employed to predict the inhibition activity of 87 structurally diverse pyrimidine-based derivatives as phosphodiestrae10A (PDE10A) inhibitors due to their potential in the treatment of schizophrenia. In this method, compounds in their preferred enzyme-docked conformations were utilized to derive interaction-based quantitative descriptors in order to explain reported PDE10A inhibitory activities. Multiple linear regression (MLR), artificial neural network (ANN), and least square support vector regression (LS-SVR) were exploited to developing the structure-based quantitative structure–activity relationship models. Among these models, LS-SVR model showed more satisfactory statistical parameters with regard to both internal (Rtrain = 0.951, Q2 = 0.804, RMSEtrain = 0.494) and external validation (Rtest = 0.941, RMSEtest = 0.549) test results. Information from the most relevant descriptors suggests that incorporating steric effect, electronegativity, and the number of substituted aromatic carbon correlate the activity with structural features of the studied compounds. Molecular docking analysis of the most potent inhibitor explored that hydrogen bond formation and hydrophobicity participated in the binding interaction of PDE10A complex active pocket which these findings are in line with those obtained from QSAR model. The reliability assessment of compounds predictions was checked by model applicability domain (AD) analysis.

A quantitative structure-biodegradation relationship (QSBR) approach to predict biodegradation rates of aromatic chemicals

The objective of this work was to develop a QSBR model for the prioritization of organic pollutants based on biodegradation rates from a database containing globally harmonized biodegradation tests using relevant molecular descriptors. To do this, we first categorized the chemicals into three groups (Group 1: simple aromatic chemicals with a single ring, Group 2: aromatic chemicals with multiple rings and Group3: Group 1 plus Group 2) based on molecular descriptors, estimated the first order biodegradation rate of the chemicals using rating values derived from the BIOWIN3 model, and finally developed, validated and defined the applicability domain of models for each group using a multiple linear regression approach. All the developed QSBR models complied with OECD principles for QSAR validation. The biodegradation rate in the models for the two groups (Group 2 and 3 chemicals) are associated with abstract molecular descriptors that provide little relevant practical information towards understanding the relationship between chemical structure and biodegradation rates. However, molecular descriptors associated with the QSBR model for Group 1 chemicals (R2 = 0.89, Q2loo = 0.87) provided information on properties that can readily be scrutinised and interpreted in relation to biodegradation processes. In combination, these results lead to the conclusion that QSBRs can be an alternative tool to estimate the persistence of chemicals, some of which can provide further insights into those factors affecting biodegradation.

Sustainable Business Models: A Review

During the past two decades of e-commerce growth, the concept of a business model has become increasingly popular. More recently, the research on this realm has grown rapidly, with diverse research activity covering a wide range of application areas. Considering the sustainable development goals, the innovative business models have brought a competitive advantage to improve the sustainability performance of organizations. The concept of the sustainable business model describes the rationale of how an organization creates, delivers, and captures value, in economic, social, cultural, or other contexts, in a sustainable way. The process of sustainable business model construction forms an innovative part of a business strategy. Different industries and businesses have utilized sustainable business models’ concept to satisfy their economic, environmental, and social goals simultaneously. However, the success, popularity, and progress of sustainable business models in different application domains are not clear. To explore this issue, this research provides a comprehensive review of sustainable business models literature in various application areas. Notable sustainable business models are identified and further classified in fourteen unique categories, and in every category, the progress -either failure or success- has been reviewed, and the research gaps are discussed. Taxonomy of the applications includes innovation, management and marketing, entrepreneurship, energy, fashion, healthcare, agri-food, supply chain management, circular economy, developing countries, engineering, construction and real estate, mobility and transportation, and hospitality. The key contribution of this study is that it provides an insight into the state of the art of sustainable business models in various application areas and future research directions. This paper concludes that popularity and the success rate of sustainable business models in all application domains have been increased along with the increasing use of advanced technologies.

Neonicotinoid insecticide design: molecular docking, multiple chemometric approaches, and toxicity relationship with Cowpea aphids.

Neonicotinoids are the fastest-growing class of insecticides successfully applied in plant protection, human and animal health care. The significant resistance increases led to the urgent need for alternative new neonicotinoids, with improved insecticidal activity. We performed molecular docking to describe a common binding mode of neonicotinoids into the nicotinic acetylcholine receptor, and to select the appropriate conformations to derive models. These were further used in a QSAR study employing both linear and nonlinear approaches to model the inhibitory activity against the Cowpea aphids. Linear modeling was performed by multiple linear regression and partial least squares and nonlinear modeling by artificial neural networks and support vector machine methods. The OECD principles were considered for QSAR models validation. Robust models with predictive power were found for neonicotinoid diverse structures. Based on our QSAR and docking outcomes, five new insecticides were predicted, according to the model applicability domain, the ligand efficiencies, and the binding mode. Therefore, the developed models can be confidently used for the prediction of the insecticidal activity of new chemicals, saving a substantial amount of time and money and, also, contributing to the chemical risk assessment.

Chemometric modeling of Daphnia magna toxicity of agrochemicals

Over the past few years, the ecotoxicological hazard potential of agrochemicals has received much attention in the industries and regulatory agencies. In the current work, we have developed quantitative structure-activity relationship (QSAR) models for Daphnia magna toxicities of different classes of agrochemicals (fungicides, herbicides, insecticides and microbiocides) individually as well as for the combined set with the application of Organization for Economic Co-operation and Development (OECD) recommended guidelines. The models for the individual data sets as well as for the combined set were generated employing only simple and interpretable two-dimensional descriptors, and subsequently strictly validated using test set compounds. The validated individual models were used to generate consensus models, with the objective to improve the prediction quality and reduced prediction errors. All the individual models of different classes of agrochemicals as well as the global set of agrochemicals showed encouraging statistical quality and prediction ability. The general observations from the derived models suggest that the toxicity increases with lipophilicity and decreases with polarity. The generated models of different classes of agrochemicals and also for the combined set should be applicable for data gap filling for new or untested agrochemical compounds falling within the applicability domain of the developed models.

Quantitative structure-property relationship studies for the prediction of the vapor pressure of volatile organic compounds

A theoretical model (QSPR) using multiple linear regression analysis for predicting the vapor pressure (pv) of volatile organic compounds (VOCs) has been developed. A series of 51 compounds were analyzed by multiple linear regression analysis. First, the data set was separated arbitrarily into a training set (39 chemicals) and a test set (12 chemicals) for statistical external validation. A four-dimensional model was developed using as independent variables theoretical descriptors derived from Dragon software when applying the GA (genetic algorithm)?VSS (variable subset selection) procedure. The obtained model was used to predict the vapor pressure of the test set compounds, and an agreement between experimental and predicted values was verified. This model, with high statistical significance (R2 = 0.9090, Q2 LOO = 0.8748, Q2 ext = 0.8307, s = 0.24), could be used adequately for the prediction and description of the log pv value of other VOCs. The applicability domain of MLR model was investigated using a William?s plot to detect outliers and outsides compounds.

Development of classification models for predicting chronic toxicity of chemicals to Daphnia magna and Pseudokirchneriella subcapitata

ABSTRACTBoth the acute toxicity and chronic toxicity data on aquatic organisms are indispensable parameters in the ecological risk assessment priority chemical screening process (e.g. persistent, bioaccumulative and toxic chemicals). However, most of the present modelling actions are focused on developing predictive models for the acute toxicity of chemicals to aquatic organisms. As regards chronic aquatic toxicity, considerable work is needed. The major objective of the present study was to construct in silico models for predicting chronic toxicity data for Daphnia magna and Pseudokirchneriella subcapitata. In the modelling, a set of chronic toxicity data was collected for D. magna (21 days no observed effect concentration (NOEC)) and P. subcapitata (72 h NOEC), respectively. Then, binary classification models were developed for D. magna and P. subcapitata by employing the k-nearest neighbour method (k-NN). The model assessment results indicated that the obtained optimum models had high accuracy, sensitivity and specificity. The model application domain was characterized by the Euclidean distance-based method. In the future, the data gap for other chemicals within the application domain on their chronic toxicity for D. magna and P. subcapitata could be filled using the models developed here.

Bimolecular Nucleophilic Substitution Reactions: Predictive Models for Rate Constants and Molecular Reaction Pairs Analysis.

Here, we report the data visualization, analysis and modeling for a large set of 4830 SN 2 reactions the rate constant of which (logk) was measured at different experimental conditions (solvent, temperature). The reactions were encoded by one single molecular graph - Condensed Graph of Reactions, which allowed us to use conventional chemoinformatics techniques developed for individual molecules. Thus, Matched Reaction Pairs approach was suggested and used for the analyses of substituents effects on the substrates and nucleophiles reactivity. The data were visualized with the help of the Generative Topographic Mapping approach. Consensus Support Vector Regression (SVR) model for the rate constant was prepared. Unbiased estimation of the model's performance was made in cross-validation on reactions measured on unique structural transformations. The model's performance in cross-validation (RMSE=0.61 logk units) and on the external test set (RMSE=0.80) is close to the noise in data. Performances of the local models obtained for selected subsets of reactions proceeding in particular solvents or with particular type of nucleophiles were similar to that of the model built on the entire set. Finally, four different definitions of model's applicability domains for reactions were examined.

Validation of the performance of TIMES genotoxicity models with EFSA pesticide data.

This study validates the performance of the TIssue MEtabolism Simulator (TIMES) genotoxicity models with data on pesticide chemicals included in a recently released European Food Safety Authority (EFSA) genotoxicity database. The EFSA database is biased towards negative chemicals. A comparison of substances included in the EFSA database and TIMES genotoxicity databases showed that the majority of the EFSA pesticides is not included in the TIMES genotoxicity databases and, thus, out of the applicability domains of the current TIMES models. However, the EFSA genotoxicity database provides an opportunity to expand the TIMES models. Where there is overlap of substances, consistency between EFSA and TIMES databases for the chemicals with documented data is found to be high (>80%) with respect to the Ames data and lower than the Ames data with respect to chromosomal aberration (CA) and mouse lymphoma assay (MLA) data. No conclusion for consistency with respect to micronucleus test and comet genotoxicity data can be provided due to the limited number of overlapping substances. Specificity of the models is important, given the prevalence of negative genotoxicity data in the EFSA database. High specificity (>80%) is obtained for prediction of the EFSA pesticides with Ames data. Moreover, this high specificity of the TIMES Ames models is not dependant on pesticides being within the domains. Specificity of the TIMES CA and MLA models is lower (>40%) to pesticides for out of domain. Sensitivity of TIMES in vitro and in vivo models cannot be properly estimated due to the small number of positive chemicals in the EFSA database.

A novel modeling domain application

This paper presents a novel to model application domain. Application domain description precedes requirements engineering, and is the basis for the development of a software or information system that satisfies all expectations of its users. The domain model is used to generate project specific process models. Our aim is to develop a model description for processes which permits to create comprehensive scenarios. Modeling can be divided into a structural, and behavioral. This paper projects that an important future direction in software engineering is domain-specific software engineering. From requirements specification to design, and then implementation, a tighter coupling between the descriptions of a software system with its application domain has the potential to improve both the correctness and reliability of the software system. The greatest challenge in this area is the evolution of the application domain itself. We show how the application domain description can be mapped to requirements and discuss engineering of application domain descriptions.