LightGBM: An Effective and Scalable Algorithm for Prediction of Chemical Toxicity–Application to the Tox21 and Mutagenicity Data Sets

Abstract

Machine learning algorithms have attained widespread use in assessing the potential toxicities of pharmaceuticals and industrial chemicals because of their faster-speed and lowercost compared to experimental bioassays. Gradient boosting is an effective algorithm that often achieves high predictivity, but historically the relative long computational time limited its applications in predicting large compound libraries or developing in silico predictive models that require frequent retraining. LightGBM, a recent improvement of the gradient boosting algorithm inherited its high predictivity but resolved its scalability and long computational time by adopting leaf-wise tree growth strategy and introducing novel techniques. In this study, we compared the predictive performance and the computational time of LightGBM to deep neural networks, random forests, support vector machines, and XGBoost. All algorithms were rigorously evaluated on publicly available Tox21 and mutagenicity datasets using a Bayesian optimization integrated nested 10-fold cross-validation scheme that performs hyperparameter optimization while examining model generalizability and transferability to new data. The evaluation results demonstrated that LightGBM is an effective and highly scalable algorithm offering the best predictive performance while consuming significantly shorter computational time than the other investigated algorithms across all Tox21 and mutagenicity datasets. We recommend LightGBM for applications in in silico safety assessment and also in other areas of cheminformatics to fulfill the ever-growing demand for accurate and rapid prediction of various toxicity or activity related endpoints of large compound libraries present in the pharmaceutical and chemical industry.

Multi‐institutional dose‐segmented dosiomic analysis for predicting radiation pneumonitis after lung stereotactic body radiation therapy

“…To develop significantly accurate models and prevent overfitting to the training datasets, hyperparameters for LightGBM were optimized using stratified fivefold cross-validation with Bayesian optimization. 29 Suitable hyperparameters were determined by searching the largest average area under the curve value in the receiver operating characteristic curve (ROC-AUC) of the five models. Furthermore, each of the final three models with the closest ROC-AUC value to the average ROC-AUC value in fivefold crossvalidation was applied to the test datasets.…”

A Toxicity Prediction Tool for Potential Agonist/Antagonist Activities in Molecular Initiating Events Based on Chemical Structures

“…Therefore, a growing interest exists in a comprehensive in silico approach to detect the potential toxicity of chemicals. The literature presents the results of successful examples of alternative in silico toxicity screening methods and their applications using the Tox21 10K library [ 19 , 20 , 21 ]. However, even though there are 59 types of well-confirmed assay results of agonist/antagonist activities for toxicity targets in the Tox21 10K library, several studies have built models for only a small number of toxicity targets.…”

Identifying sarcopenia in advanced non‐small cell lung cancer patients using skeletal muscle CT radiomics and machine learning

“…In order to address this issue, it is advisable to adopt Bayesian optimization strategies to perform hyperparameter tuning, as it has better performance on the test set and requires fewer iterations compared to grid and random searches. 29,30 Our results demonstrated that the baseline lightGBM model could be improved via Bayesian optimization by over 8.0% in ACC value, 11.1% in F-value, 21.9% in MCC value, and 8.5% in AUC value, on the validation set, respectively, thereby verifying the effectiveness of Bayesian optimization.…”