The ability to assess acute oral toxicity accurately is necessary for ensuring the safety of chemicals. One of the most effective ways to enhance computational predictions of acute oral toxicity is by leveraging the power of large databases of experimental data. This approach improves the accuracy of in silico models, reducing the need for additional animal testing. Here, we explore how data abundance and mechanistic understanding interplay to advance acute oral toxicity predictions and introduce new and updated models designed to predict GHS and CLP categories.
Data Abundance plays a pivotal role in the performance of in silico models. When large volumes of high-quality data are available, statistical models can identify patterns and make accurate predictions even if the underlying mechanisms are not fully understood. This is particularly true for acute rat oral toxicity, where the mechanistic understanding leading to toxicity may be limited, but extensive datasets from various studies and standardized testing methods provide a wealth of information. With a database of 20,000 entries, including lethal doses (LD50), the robustness of predictive models is significantly enhanced.
Mechanistic Understanding is another critical factor. In general, when both data abundance and mechanistic understanding are high, models perform exceptionally well, leveraging the strengths of both aspects. Acute oral toxicity can result from various mechanisms, but the complete range of these mechanisms and the structure-activity relationships that govern them are not fully understood. This complexity requires an extensive training database.
By leveraging large databases, we can develop more accurate and reliable in silico models. As we continue to refine these models and explore new endpoints, the interplay between data and mechanisms summarized in Figure 1 is central to our work. The development of our acute toxicity models to predict GHS and CLP categories marks a significant step forward in this ongoing effort.
The new acute toxicity databases and models will be made available in our 2025 release. The new acute toxicity suite will highlight:
- Addition of thousands of new records for acute toxicity data.
- Experimental data are now categorized based on two categorization schemes: one predicting CLP categories and the other GHS categories.
- Predictive models support the derivation of acute rat oral GHS categories and CLP categories.
- New alerts and statistical models predicting CLP categories for Acute Rat Oral, Acute Mouse Oral, Acute Rat Dermal, and Acute Rabbit Dermal Toxicity.
Figure 1: Interplay between data and mechanisms. Note that mechanistic understanding is a continuous scale and depth of knowledge of mechanisms differs across chemical classes of compounds.
Please contact us to speak to an expert today or learn more about our solutions here.
