Science Inventory

In vitro transcriptional points of departure from human whole transcriptome and surrogate transcriptome (S1500+) targeted RNA-Seq assays are highly comparable

Citation:

Johnson, J., J. Bundy, J. Harrill, R. Judson, AND L. Everett. In vitro transcriptional points of departure from human whole transcriptome and surrogate transcriptome (S1500+) targeted RNA-Seq assays are highly comparable. 2024 Society of Toxicology Annual Meeting, Salt Lake City, UT, March 10 - 14, 2024. https://doi.org/10.23645/epacomptox.25533505

Impact/Purpose:

This is an abstract for a poster presentation being submitted to the 2024 Society of Toxicology meeting.

Description:

Background               The EPA has been investigating high-throughput screening methods using in vitro technologies to reduce the use of costly and time-consuming in vivo animal models for toxicological testing. The EPA has performed high-throughput transcriptomic screening on over 1,000 chemicals in three different human cell lines and used a signature concentration-response analysis method (Harrill, et al. Toxicol Sci 2021) to summarize the biological pathways perturbed by each chemical exposure, the results of which are available on the CompTox Chemicals Dashboard [comptox.epa.gov/dashboard]. These screening efforts have relied on a whole transcriptome profiling assay (Yeakley, et al. PLoS ONE 2017) that measures the expression of over 20,000 genes; however, a reduced representation [PFK(1] [PFK(2] [JJ3] version of the assay, called “S1500+” (Mav, et al. PLoS ONE 2018) is also available and can be run at a lower cost. Here, we investigated the potential impacts of using the S1500+ reduced coverage assay in lieu of the whole transcriptome assay for calculating potency estimates and identifying perturbed biological pathways. Methods               In order to test whether the S1500+ assay could provide similar results and conclusions as the whole transcriptome [PFK(4] [JJ5] [EL(6] assay, we re-analyzed the large-scale screening data sets for 3 human cell lines (U-2 OS, HepaRG, and MCF7) using only the expression measurements for ~2,700 genes captured by the S1500+ assay. These data were transformed into gene signature scores and then used for concentration-response modeling, similar to how the whole transcriptome data was previously analyzed. Signature-level concentration-response models based on both the whole transcriptome and S1500+ gene sets were then filtered to remove inactive signatures (top/cutoff < 1 or hitcall < 0.9), [PFK(7] [JJ8] [EL(9] and an overall transcriptomic point of departure (tPOD) was derived for each chemical sample, cell line, and underlying gene set. We then compared the corresponding log10-transformed tPODs from whole transcriptome and S1500+-based analyses to examine the overall concordance of these values. These log10 tPODs were tested to see if they were within +/- 1 order of magnitude and if a univariate linear model could predict the relationship between the whole transcriptome gene signatures and the S1500+ gene signatures. Results               The tPODs derived from only the S1500+ genes were highly concordant with the tPODs derived from the whole transcriptome expression data. Regardless of cell type, the relationship between the two sets of log10 tPODs appears to follow a tight normal distribution along an identity line (slope of 1 and intercept of 0). The Spearman correlation coefficients of log10 tPODs between assay platforms for each of the cell types are between 0.66 to 0.75, and over 97% of the tPODs fall within 1 order of magnitude from the identity line. [PFK(10] [JJ11] [EL(12] These results were based on over 1,200 chemicals profiled in 3 distinct cell lines with a wide range of tPODs spanning 5 orders of magnitude, demonstrating the robustness and broad applicability of these results. Surprisingly, the total number of active gene signatures only decreased by 10% to 26%, depending on cell type, when reducing the analysis to the S1500+ genes, despite the fact that the S1500+ genes only account for approximately 10% of all genes tested in the whole transcriptome assay. Conclusion                 These results demonstrate that despite the differences in gene response across a diverse set of chemical exposures and cell types, the current signature-based approach still provides consistent tPODs even when including a fraction of the measured genes. Furthermore, running only the genes on the S1500+ platform appears to still allow for concentration-response analysis of the majority of active signatures identified when running the whole , and with minimal changes in overall

Record Details:

Record Type:DOCUMENT( PRESENTATION/ POSTER)
Product Published Date:03/14/2024
Record Last Revised:04/03/2024
OMB Category:Other
Record ID: 360990