Hongbin Yang

The applicability domain of machine learning models trained on structural fingerprints for the prediction of biological endpoints is often limited by the lack of diversity of chemical space of the training data. In this work, we developed similarity-based merger models which combined the outputs of individual models trained on cell morphology (based on Cell Painting) and chemical structure (based on chemical fingerprints) and the structural and morphological similarities of the compounds in the test dataset to compounds in the training dataset. We applied these similarity-based merger models using logistic regression models on the predictions and similarities as features and predicted assay hit calls of 177 assays from ChEMBL, PubChem and the Broad Institute (where the required Cell Painting annotations were available). We found that the similarity-based merger models outperformed other models with an additional 20% assays (79 out of 177 assays) with an AUC>0.70 compared with 65 out of 177 assays using structural models and 50 out of 177 assays using Cell Painting models. Our results demonstrated that similarity-based merger models combining structure and cell morphology models can more accurately predict a wide range of biological assay outcomes and further expanded the applicability domain by better extrapolating to new structural and morphology spaces. Abstract Figure: Figure: For TOC Only

Fish is one of the model animals used to evaluate the adverse effects of a chemical exposed to the ecosystem. However, its low throughput and relevantly high expense make it impossible to test all new chemicals in manufacture. Hence, using in silico models to prioritize compounds to be tested has been widely applied in environmental risk assessment and drug discovery. In this study, we constructed the local predictive models for four fish species, including bluegill sunfish, rainbow trout, fathead minnow, and sheepshead minnow, and the global models with all four fish data. A total of 1874 unique compounds with their labels, that is, toxic (LC₅₀ < 10 ppm) or nontoxic, were collected from ECOTOX and literature. Both conventional machine learning methods and the deep learning architecture, graph convolutional network (GCN), were used to build predictive models. The classification accuracy of the best local model for each fish species was higher than 0.83. For the global models, two strategies including consistency prediction and probability threshold were adopted to improve the predictive capability at the cost of limiting applicability domain. For 63% of compounds in domain, the accuracy was around 0.97. By comparison of the deep learning and machine learning methods, we found that the single‐task GCN showed specific advantages in performance, and multitask GCN showed no advantages over the conventional machine learning methods. The data and models are available on GitHub (https://github.com/ChemPredict/ChemicalAquaticToxicity).

The prediction and optimization of pharmacokinetic properties are essential in lead optimization. Traditional strategies mainly depend on the empirical chemical rules from medicinal chemists. However, with the rising amount of data, it is getting more difficult to manually extract useful medicinal chemistry knowledge. To this end, we introduced IDL-PPBopt, a computational strategy for predicting and optimizing the plasma protein binding (PPB) property based on an interpretable deep learning method. At first, a curated PPB data set was used to construct an interpretable deep learning model, which showed excellent predictive performance with a root mean squared error of 0.112 for the entire test set. Then, we designed a detection protocol based on the model and Wilcoxon test to identify the PPB-related substructures (named privileged substructures, PSubs) for each molecule. In total, 22 general privileged substructures (GPSubs) were identified, which shared some common features such as nitrogen-containing groups, diamines with two carbon units, and azetidine. Furthermore, a series of second-level chemical rules for each GPSub were derived through a statistical test and then summarized into substructure pairs. We demonstrated that these substructure pairs were equally applicable outside the training set and accordingly customized the structural modification schemes for each GPSub, which provided alternatives for the optimization of the PPB property. Therefore, IDL-PPBopt provides a promising scheme for the prediction and optimization of the PPB property and would be helpful for lead optimization of other pharmacokinetic properties.

Mitochondrial toxicity is an important safety endpoint in drug discovery. Models based solely on chemical structure for predicting mitochondrial toxicity are currently limited in accuracy and applicability domain to the chemical space of the training compounds. In this work, we aimed to utilize both -omics and chemical data to push beyond the state-of-the-art. We combined Cell Painting and Gene Expression data with chemical structural information from Morgan fingerprints for 382 chemical perturbants tested in the Tox21 mitochondrial membrane depolarization assay. We observed that mitochondrial toxicants significantly differ from non-toxic compounds in morphological space and identified compound clusters having similar mechanisms of mitochondrial toxicity, thereby indicating that morphological space provides biological insights related to mechanisms of action of this endpoint. We further showed that models combining Cell Painting, Gene Expression features and Morgan fingerprints improved model performance on an external test set of 236 compounds by 60% (in terms of F1 score) and improved extrapolation to new chemical space. The performance of our combined models was comparable with dedicated in vitro assays for mitochondrial toxicity; and they were able to detect mitochondrial toxicity where Tox21 assays outcomes were inconclusive because of cytotoxicity. Our results suggest that combining chemical descriptors with different levels of biological readouts enhances the detection of mitochondrial toxicants, with practical implications for use in drug discovery.

Cell morphology features, such as those from the Cell Painting assay, can be generated at relatively low costs and represent versatile biological descriptors of a system and thereby compound response. In this study, we explored cell morphology descriptors and molecular fingerprints, separately and in combination, for the prediction of cytotoxicity- and proliferation-related in vitro assay endpoints. We selected 135 compounds from the MoleculeNet ToxCast benchmark data set which were annotated with Cell Painting readouts, where the relatively small size of the data set is due to the overlap of required annotations. We trained Random Forest classification models using nested cross-validation and Cell Painting descriptors, Morgan and ErG fingerprints, and their combinations. While using leave-one-cluster-out cross-validation (with clusters based on physicochemical descriptors), models using Cell Painting descriptors achieved higher average performance over all assays (Balanced Accuracy of 0.65, Matthews Correlation Coefficient of 0.28, and AUC-ROC of 0.71) compared to models using ErG fingerprints (BA 0.55, MCC 0.09, and AUC-ROC 0.60) and Morgan fingerprints alone (BA 0.54, MCC 0.06, and AUC-ROC 0.56). While using random shuffle splits, the combination of Cell Painting descriptors with ErG and Morgan fingerprints further improved balanced accuracy on average by 8.9% (in 9 out of 12 assays) and 23.4% (in 8 out of 12 assays) compared to using only ErG and Morgan fingerprints, respectively. Regarding feature importance, Cell Painting descriptors related to nuclei texture, granularity of cells, and cytoplasm as well as cell neighbors and radial distributions were identified to be most contributing, which is plausible given the endpoint considered. We conclude that cell morphological descriptors contain complementary information to molecular fingerprints which can be used to improve the performance of predictive cytotoxicity models, in particular in areas of novel structural space.

Structural alerts are a simple and easy way to identify toxic compounds being widely used in environmental toxicology research and drug discovery. With the emergence of big data techniques in recent years and their applications in chemistry and toxicology, computational approaches have become a promising method to identify structural alerts. In this Review, we describe the recent progress in computational methods for identification of structural alerts and their applications in toxicology. Two major computational approaches, namely frequency analysis and interpretable machine learning models, are reviewed. Recent studies have shown that both approaches are superior to expert systems with respect to predictive capability. Methodologies for defining the applicability domain of such approaches are also reviewed, with their importance stemming from their ability to not only improve the predictive performance of structural alert models but also ensure the confidence of a prediction. In addition to toxicity prediction, structural alerts could be also used to explain quantitative structure–activity relationship models and guide lead optimization in drug discovery. Nevertheless, there are still some challenges to be solved, such as how to address the co-existence of several structural alerts in one molecule, how to directly compare computationally derived structural alerts with expert systems, and how to explore new mechanisms of toxicity.

Renal clearance (CLr) plays an essential role in the elimination of drugs. In this study, 636 compounds were obtained from various sources to develop in silico models for the prediction of CLr. Step-wise multiple linear regression and random forest regression methods were employed to build global models and local models according to ionization state or net elimination pathways. The local models towards compounds undergoing different net elimination pathways showed good predictive power: the geometric mean fold error was close to 2, indicating the clearance of most compounds could be predicted within a two-fold error range. Six classification methods were used to construct classification models. However, the performance of these classification models was less than satisfactory, and the mean accuracy of the top five models in test sets was 0.65. Moreover, qualitative analysis of physicochemical profiles between compounds undergoing different net elimination pathways revealed that compounds with higher lipophilicity tended to be reabsorbed more easily and showed lower CLr, while compounds with higher values of polar descriptors tended to secrete more easily and showed higher CLr.