You are here:
Transitioning Generalized Read-across (GenRA) towards quantitative predictions (SOT 2020)
Citation:
Helman, G., I. Shah, AND G. Patlewicz. Transitioning Generalized Read-across (GenRA) towards quantitative predictions (SOT 2020). Society of Toxicology 59th Annual Meeting 2020, Anaheim, CA, March 15 - 19, 2020. https://doi.org/10.23645/epacomptox.17139149
Impact/Purpose:
Presentation to the Society of Toxicology 59th Annual Meeting March 2020. The development of novel artificial intelligence approaches based on massive public toxicity data is urgently needed to generate new predictive models for chemical toxicity evaluations and establish scientific confidence in the developed models as alternatives for evaluating untested compounds. In this procedure, traditional approaches (e.g., QSAR) purely based on chemical structures have been replaced by newly designed data-driven and mechanism-driven modeling. The resulting models realize the concept of Adverse Outcome Pathway (AOP), which can not only directly evaluate toxicity potentials of new compounds but also illustrate relevant toxicity mechanisms. The recent advancements of computational toxicology in the big data era are paving the road to future toxicity testing, and will have significant impacts on public health.
Description:
Computational approaches have recently gained popularity in the field of read-across to automatically fill data-gaps for untested chemicals. To this end, we developed a generalized read-across (GenRA) approach, which utilizes chemical descriptor information and/or in vitro bioactivity data to qualitatively predict hazard based on study data summarized in US’s EPA ToxRefDB v1.0. Following the success in qualitatively predicting hazard by study type and target organ or system, a clear need surfaced: quantitative predictions of the potency of possible hazard based on analogous chemicals. To address this need, the GenRA workflow was modified to evaluate quantitative predictions using Morgan chemical fingerprints with a minimum similarity threshold of 0.5 and a maximum of 10 nearest neighbors in 2 case studies: 1) LD50 values using a large dataset of rate oral acute toxicity (LD50 values for 7011 discrete chemicals) and 2) point of departure (POD) values (oral Lowest Observed Adverse Effect Levels (LOAEL) for 1,076 chemicals) obtained from ToxRefDB v2.0. An R2 value of 0.61 and RMSE of 0.58 was achieved for GenRA LD50 predictions. Cross validated R2 values ranged from 0.52 - 0.69. Average R2 values improved up to 0.91 for GenRA LD50 predictions within structural clusters. R2 values of 0.43, 0.22, 0.14, and 0.26, were achieved for systemic, developmental, reproductive, and cholinesterase LOAELs respectively. Average R2 values for systemic, developmental, reproductive, and cholinesterase LOAEL predictions improved to 0.73, 0.66, 0.60 and 0.79 for structural clusters. Our findings highlight the complexity of the chemical-toxicity landscape and the importance of identifying local domains where GenRA can be used effectively for predicting quantitative toxicity values. This abstract does not reflect EPA policy.
URLs/Downloads:
DOI: Transitioning Generalized Read-across (GenRA) towards quantitative predictions (SOT 2020)
GENRA_SOT_PATLEWICZ_2020.PDF (PDF, NA pp, 1805.783 KB, about PDF)