In silico methods are essential in the safety assessment of chemicals. Compared to animal testing, they can detect the possibility of an unwanted event early on, are more time-and cost-efficient, and are less connected to ethical concerns. Computer-based risk assessment offers several approaches, data science-based methods being an increasingly important sub-group. One of the substantial benefits of data science is that it allows using existing data to find correlations, build strong hypotheses, and create new, usable knowledge. This can make some resource-intensive experiments obsolete, contributing to the sustainability of the approaches. In toxicity modeling, two of the essential factors to consider in choosing the suitable method are the available data and desired toxicity endpoint. The complexity of the endpoint can impact the success rate of the machine learning model. It is possibly not enough to look at the toxicity based on structural descriptors of the compounds yet view the toxic event more systematically. We propose a data science pipeline that connects compounds to their tissue-specific biological targets and pathways. Based on this information, predictive models were built, addressing the hepatotoxicity of drug-like compounds. Several models were trained, the combination of target and pathway level descriptors and the Random Forest Algorithm providing the best performance. After building the models, descriptors for model performance were addressed and examined for a biological explanation to define which targets or pathways can have a crucial role in toxicity. Descriptors connected to some Cytochromes P450 enzymes and Biological Oxidation were ranked high. Furthermore, the involvement of other less discussed processes on connection to toxicity, such as the Biosynthesis of specialized pro-resolving mediators, was marked as fundamental. In combination with the chemical properties of the compounds, the pipeline can create a valuable tool in predicting toxicity.

Data science pipeline for using the systemic effect of drugs in the prediction of toxicity

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.