To produce functional materials with desired qualities, in silico prediction of physio-chemical characteristics of materials’ components has become an essential tool in modern chemists' toolbox. In this work, we propose a Quantitative Structure-Activity Relationship (QSAR) model that estimates toxicity for a large set of nitroaromatic compounds. The model's theoretical descriptors were derived from the structure of the compounds and selected from an initially generated set of 4500 descriptors. A super learner model was created with the greatest predictive performance coming from two Support Vector Regression (SVR) base models followed by linear regression to fine-tune the out-of-SVR predictions. Various methods were used to validate the predictability and robustness of the super learner, including squared correlation coefficient (R2), Root Means Square Error (RMSE), mean absolute error (MAE) on the training set, test set, and true external test set. Finally, this research addresses the contribution of each descriptor to the toxicity using the Accumulated Local Effect (ALE) approach.

In silico prediction of the toxicity of nitroaromatic compounds: Application of super learner QSAR approach

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.