You are here:
Confidence in Fitting and Hitting Concentration-Response Data: Tox21 10k Library Pipeline Comparison
Citation:
Sipes, N., R. Huang, K. Shockley, M. Martin, A. Shapiro, J. Addington, S. Auerbach, R. Paules, R. Judson, K. Houck, H. Hong, AND J. Hsieh. Confidence in Fitting and Hitting Concentration-Response Data: Tox21 10k Library Pipeline Comparison. Presented at Society of Toxicology Annual Meeting, San Antonio, TX, March 11 - 15, 2018. https://doi.org/10.23645/epacomptox.6945911
Impact/Purpose:
Our comparative analysis provides the scientific community inclusive access and evaluation of Tox21 data with the ability to identify higher confidence activity calls across pipelines. Poster presentation at Society of Toxicology Annual Meeting March 2018
Description:
The Tox21 program has generated high-throughput screening data on thousands of chemicals. While the data are publicly available through partner websites, PubChem, and publications, the analyses are different. We developed a pipeline consensus to identify higher confidence chemical-assay calls and are developing a public web application. Tox21 chemical-assay pair activity calls (active, inactive and, in some cases, inconclusive) were compared among the 4 hit-call methods; CurveP and 3Stage from NIEHS, TCPL from the US EPA, and CurveClass from NCATS. Out of the 664,463 chemical-assay pairs (8,948 chemicals, 67 assays), 82% had total agreement (97% inactive), 13% had 3 pipelines in agreement where the agreement call was inactive (50%), active (23%), and inconclusive (27%), 4% had a 50/50 split, and the rest at 1%. High agreement assays were nuclear receptor agonist assays (e.g., androgen, thyroid, estrogen). Complete curves with high efficacies were overly represented in these assay curve fits. Antagonist assays were over represented in the assays with the lowest total agreement, this was expected since not all pipelines use the viability counter screen to adjust the calls. In addition, the lowest total agreement was associated with higher discordance among significance thresholds. Chemical purity did not appear to be an influential factor. Chemicals with the highest positive agreement included metals (e.g., phenylmercuric acetate, zinc pyrithione, and tributyltin chloride) occurring between 70-100 times, while chemicals with the lowest agreement (e.g., cycloheximide, fulvestrant and triclocarban) occurred about 15 times. It is important to note that some chemicals are more concordant in specific assays than others. In addition, median differences between calculated maximum efficacies and the pipeline’s predetermined minimal activity significance threshold were lower when calls were discordant versus when all four pipelines agreed on an active call. Our comparative analysis provides the scientific community inclusive access and evaluation of Tox21 data with the ability to identify higher confidence activity calls across pipelines. This abstract does not represent any US government policy.
URLs/Downloads:
DOI: Confidence in Fitting and Hitting Concentration-Response Data: Tox21 10k Library Pipeline Comparison
SOT2018_NSS_TOX21HITCC_02272018_FINAL.PDF (PDF, NA pp, 1670.585 KB, about PDF)