Prediction of skin sensitization with a particle swarm optimized support vector machine

Development of a hierarchical support vector regression-based in silico model for the prediction of the cysteine depletion in DPRA

Topical and transdermal treatments have been dramatically growing recently and it is crucial to consider skin sensitization during the drug discovery and development process for these administration routes. Various tests, including animal and non-animal approaches, have been devised to assess the potential for skin sensitization. Furthermore, numerous in silico models have been created, providing swift and cost-effective alternatives to traditional methods such as in vivo, in vitro, and in chemico methods for categorizing compounds. In this study, a quantitative structure-activity relationship (QSAR) model was developed using the innovative hierarchical support vector regression (HSVR) scheme. The aim was to quantitatively predict the potential for skin sensitization by analyzing the percent of cysteine depletion in Direct Peptide Reactivity Assay (DPRA). The results demonstrated accurate, consistent, and robust predictions in the training set, test set, and outlier set. Consequently, this model can be employed to estimate skin sensitization potential of novel or virtual compounds.

Prediction of skin sensitization using machine learning

As global awareness of animal welfare spreads, the development of alternative animal test models is increasingly necessary. The purpose of this study was to develop a practical machine-learning model for skin sensitization using three physicochemical properties of the chemicals: surface tension, melting point, and molecular weight. In this study, a total of 482 chemicals with local lymph node assay results were collected, and 297 datasets with 6 physico-chemical properties were used to develop Random Forest (RF) model for skin sensitization. The developed model was validated with 45 fragrance allergens announced by European Commission. The validation results showed that RF achieved better or similar classification performance with f1-scores of 54% for penal, 82% for ternary, and 96% for binary compared with Support Vector Machine (SVM) (penal, 41%; ternary, 81%; binary, 93%), QSARs (ChemTunes, 72% for ternary; OECD Toolbox, 89% for binary), and a linear model (Kim et al., 2020) (41% for penal), and we recommend the ternary classification based on Global Harmonized System providing more detailed and precise information. In the further study, the proposed model results were experimentally validated with the Direct Peptide Reactivity Assay (DPRA, OECD TG 442C approved model), and the results showed a similar tendency. We anticipate that this study will help to easily and quickly screen chemical sensitization hazards.

PreS/MD: Predictor of Sensitization Hazard for Chemical Substances Released From Medical Devices

In the United States, a pre-market regulatory submission for any medical device that comes into contact with either a patient or the clinical practitioner must include an adequate toxicity evaluation of chemical substances that can be released from the device during its intended use. These substances, also referred to as extractables and leachables, must be evaluated for their potential to induce sensitization/allergenicity, which traditionally has been done in animal assays such as the guinea pig maximization test (GPMT). However, advances in basic and applied science are continuously presenting opportunities to employ new approach methodologies, including computational methods which, when qualified, could replace animal testing methods to support regulatory submissions. Herein, we developed a new computational tool for rapid and accurate prediction of the GPMT outcome that we have named PreS/MD (predictor of sensitization for medical devices). To enable model development, we (1) collected, curated, and integrated the largest publicly available dataset for GPMT results; (2) succeeded in developing externally predictive (balanced accuracy of 70%-74% as evaluated by both 5-fold external cross-validation and testing of novel compounds) quantitative structure-activity relationships (QSAR) models for GPMT using machine learning algorithms, including deep learning; and (3) developed a publicly accessible web portal integrating PreS/MD models that can predict GPMT outcomes for any molecule of interest. We expect that PreS/MD will be used by both industry and regulatory scientists in medical device safety assessments and help replace, reduce, or refine the use of animals in toxicity testing. PreS/MD is freely available at https://presmd.mml.unc.edu/.

Transfer learning for predicting human skin sensitizers

Computational prioritization of chemicals for potential skin sensitization risks plays essential roles in the risk assessment of environmental chemicals and drug development. Given the huge number of chemicals for testing, computational methods enable the fast identification of high-risk chemicals for experimental validation and design of safer alternatives. However, the development of robust prediction model requires a large dataset of tested chemicals that is usually not available for most toxicological endpoints, especially for human data. A small training dataset makes the development of effective models difficult with insufficient coverage and accuracy. In this study, an ensemble tree-based multitask learning method was developed incorporating three relevant tasks in the well-defined adverse outcome pathway (AOP) of skin sensitization to transfer shared knowledge to the major task of human sensitizers. The results show both largely improved coverage and accuracy compared with three state-of-the-art methods. A user-friendly prediction server was available at https://cwtung.kmu.edu.tw/skinsensdb/predict . As AOPs for various toxicity endpoints are being actively developed, the proposed method can be applied to develop prediction models for other endpoints.

Shuffling cross-validation-bee algorithm as a new descriptor selection method for retention studies of pesticides in biopartitioning micellar chromatography

Bee algorithm (BA) is an optimization algorithm inspired by the natural foraging behaviour of honey bees to find the optimal solution which can be proposed to feature selection. In this paper, shuffling cross-validation-BA (CV-BA) was applied to select the best descriptors that could describe the retention factor (log k) in the biopartitioning micellar chromatography (BMC) of 79 heterogeneous pesticides. Six descriptors were obtained using BA and then the selected descriptors were applied for model development using multiple linear regression (MLR). The descriptor selection was also performed using stepwise, genetic algorithm and simulated annealing methods and MLR was applied to model development and then the results were compared with those obtained from shuffling CV-BA. The results showed that shuffling CV-BA can be applied as a powerful descriptor selection method. Support vector machine (SVM) was also applied for model development using six selected descriptors by BA. The obtained statistical results using SVM were better than those obtained using MLR, as the root mean square error (RMSE) and correlation coefficient (R) for whole data set (training and test), using shuffling CV-BA-MLR, were obtained as 0.1863 and 0.9426, respectively, while these amounts for the shuffling CV-BA-SVM method were obtained as 0.0704 and 0.9922, respectively.

A ternary classification using machine learning methods of distinct estrogen receptor activities within a large collection of environmental chemicals

Endocrine-disrupting chemicals (EDCs), which can threaten ecological safety and be harmful to human beings, have been cause for wide concern. There is a high demand for efficient methodologies for evaluating potential EDCs in the environment. Herein an evaluation platform was developed using novel and statistically robust ternary models via different machine learning models (i.e., linear discriminant analysis, classification and regression tree, and support vector machines). The platform is aimed at effectively classifying chemicals with agonistic, antagonistic, or no estrogen receptor (ER) activities. A total of 440 chemicals from the literature were selected to derive and optimize the three-class model. One hundred and nine new chemicals appeared on the 2014 EPA list for EDC screening, which were used to assess the predictive performances by comparing the E-screen results with the predicted results of the classification models. The best model was obtained using support vector machines (SVM) which recognized agonists and antagonists with accuracies of 76.6% and 75.0%, respectively, on the test set (with an overall predictive accuracy of 75.2%), and achieved a 10-fold cross-validation (CV) of 73.4%. The external predicted accuracy validated by the E-screen assay was 87.5%, which demonstrated the application value for a virtual alert for EDCs with ER agonistic or antagonistic activities. It was demonstrated that the ternary computational model could be used as a faster and less expensive method to identify EDCs that act through nuclear receptors, and to classify these chemicals into different mechanism groups.

SkinSensDB: a curated database for skin sensitization assays

Skin sensitization is an important toxicological endpoint for chemical hazard determination and safety assessment. Prediction of chemical skin sensitizer had traditionally relied on data from rodent models. The development of the adverse outcome pathway (AOP) and associated alternative in vitro assays have reshaped the assessment of skin sensitizers. The integration of multiple assays as key events in the AOP has been shown to have improved prediction performance. Current computational models to predict skin sensitization mainly based on in vivo assays without incorporating alternative in vitro assays. However, there are few freely available databases integrating both the in vivo and the in vitro skin sensitization assays for development of AOP-based skin sensitization prediction models. To facilitate the development of AOP-based prediction models, a skin sensitization database named SkinSensDB has been constructed by curating data from published AOP-related assays. In addition to providing datasets for developing computational models, SkinSensDB is equipped with browsing and search tools which enable the assessment of new compounds for their skin sensitization potentials based on data from structurally similar compounds. SkinSensDB is publicly available at http://cwtung.kmu.edu.tw/skinsensdb.

QSAR models of human data can enrich or replace LLNA testing for human skin sensitization

Skin sensitization is a major environmental and occupational health hazard. Although many chemicals have been evaluated in humans, there have been no efforts to model these data to date. We have compiled, curated, analyzed, and compared the available human and LLNA data. Using these data, we have developed reliable computational models and applied them for virtual screening of chemical libraries to identify putative skin sensitizers. The overall concordance between murine LLNA and human skin sensitization responses for a set of 135 unique chemicals was low (R = 28-43%), although several chemical classes had high concordance. We have succeeded to develop predictive QSAR models of all available human data with the external correct classification rate of 71%. A consensus model integrating concordant QSAR predictions and LLNA results afforded a higher CCR of 82% but at the expense of the reduced external dataset coverage (52%). We used the developed QSAR models for virtual screening of CosIng database and identified 1061 putative skin sensitizers; for seventeen of these compounds, we found published evidence of their skin sensitization effects. Models reported herein provide more accurate alternative to LLNA testing for human skin sensitization assessment across diverse chemical data. In addition, they can also be used to guide the structural optimization of toxic compounds to reduce their skin sensitization potential.

Can mutagenicity information be useful in an Integrated Testing Strategy (ITS) for skin sensitization?

Our previous work has investigated the utility of mutagenicity data in the development and application of Integrated Testing Strategies (ITS) for skin sensitization by focusing on the chemical mechanisms at play and substantiating these with experimental data where available. The hybrid expert system TIMES (Tissue Metabolism Simulator) was applied in the identification of the chemical mechanisms since it encodes a comprehensive set of established structure-activity relationships for both skin sensitization and mutagenicity. Based on the evaluation, the experimental determination of mutagenicity was thought to be potentially helpful in the evaluation of skin sensitization potential. This study has evaluated the dataset reported by Wolfreys and Basketter (Cutan. Ocul. Toxicol. 23 (2004), pp. 197-205). Upon an update of the experimental data, the original reported concordance of 68% was found to increase to 88%. There were several compounds that were 'outliers' in the two experimental evaluations which are discussed from a mechanistic basis. The discrepancies were found to be mainly associated with the differences between skin and liver metabolism. Mutagenicity information can play a significant role in evaluating sensitization potential as part of an ITS though careful attention needs to be made to ensure that any information is interpreted in the appropriate context.

Prediction of skin sensitization potential using D-optimal design and GA-kNN classification methods

Modelling of skin sensitization data of 255 diverse compounds and 450 calculated descriptors was performed to develop global predictive classification models that are applicable to whole chemical space. With this aim, we employed two automated procedures, (a) D-optimal design to select optimal members of the training and test sets and (b) k-Nearest Neighbour classification (kNN) method along with Genetic Algorithms (GA-kNN Classification) to select significant and independent descriptors in order to build the models. This methodology helped us to derive multiple models, M1-M5, that are stable and robust. The best among them, model M1 (CCR(train) = 84.3%, CCR(test) = 87.2% and CCR(ext) = 80.4%), is based on six neighbours and nine descriptors and further suggests that: (a) it is stable and robust and performs better than the reported models in literature, and (b) the combination of D-optimal design and GA-kNN classification approach is a very promising approach. Consensus prediction based on the models M1-M5 improved the CCR of training, test and external validation datasets by 3.8%, 4.45% and 3.85%, respectively, over M1. From the analysis of the physical meaning of the selected descriptors, it is inferred that the skin sensitization potential of small organic compounds can be accurately predicted using calculated descriptors that code for the following fundamental properties: (i) lipophilicity, (ii) atomic polarizability, (iii) shape, (iii) electrostatic interactions, and (iv) chemical reactivity.