You are here:
Profiling the ToxCast library with pluripotent embryonic stem cell assays
Citation:
Knudsen, T., T. Zurlinden, AND S. Hunter. Profiling the ToxCast library with pluripotent embryonic stem cell assays. Presented at Teratology Society Annual Meeting, Clearwater, FL, June 23 - 27, 2018. https://doi.org/10.23645/epacomptox.6949823
Impact/Purpose:
ToxCast chemicals were profiled for developmental toxicity potential in two embryonic stem cell assays and the data was processed in the ToxCast data analysis pipeline. To gain insight into the biological pathways and targets associated with the stem cell responses, machine-learning was used to mine the strongest positive and negative correlates to 337 enzymatic and receptor signaling assays in the ToxCast NovaScreen dataset (NVS). The findings of this study set the stage for identifying and developing new approach methodologies based on in vitro data and in silico models for prenatal developmental toxicity.
Description:
ToxCast chemicals were profiled for developmental toxicity potential in two embryonic stem cell assays and the data was processed in the ToxCast data analysis pipeline (tcpl): the devTOX quickPredict platform from Stemina (STM) is a human pluripotent H9 stem cell-based assay; and the mouse embryonic stem cell (mESC) adherent assay. Using the STM model, we screened 1065 ToxCast chemicals. Model performance was 81% accuracy (sensitivity 0.56, specificity 0.91) based on 171 compounds in the dataset with concordance in prenatal rat and rabbit developmental toxicity studies showing a development lowest effect level (dLEL) ≤200 mg/kg/day or no dLEL ≥ 1000 mg/kg/day) in ToxRefDB. Of the 1065 screened chemicals, the STM model predicted 181 (17% tested) as putative developmental toxicants. Using the mESC adherent assay, we screened 214 chemicals with concordance in prenantal rat or rabbit developmental toxicity studies in ToxRefDB and found model performance of 84% accuracy with 89 (42% tested) predicted putative developmental toxicants. To gain insight into the biological pathways and targets associated with the stem cell responses, machine-learning was used to mine the strongest positive and negative correlates to 337 enzymatic and receptor signaling assays in the ToxCast NovaScreen dataset (NVS). Each NVS assay was mined for an AC50 correlation against STM-positive and STM-negative results and weighted based on an assay-specific logistic regression model. Gene annotations and weighting scores for NVS targets were imported to the Reactome HSA Pathway Browser (v3.5, database release 63) to further characterize the sensitive and insensitive biochemical domains. For example, BRAF signaling was in the sensitive domain whereas estrogen signaling the insensitive domain. When the mESC response was mined against ToxCast assay targets, those correlating with developmental toxicity included TP53, RXRB, SLC18A2, CYP3A4 whereas CD38, CHRNA7, IL8 were not correlated with the mESC response. These findings set the stage for identifying and developing new approach methodologies based on in vitro data and in silico models for prenatal developmental toxicity. This abstract may not reflect US EPA policy.
URLs/Downloads:
DOI: Profiling the ToxCast library with pluripotent embryonic stem cell assays
2018-06-21_DEVTOX_TS_POSTER_V3.PDF (PDF, NA pp, 1698.527 KB, about PDF)