2013 Progress Report: Carolina Center for Computational Toxicology: Assays, models and tools for NextGen safety assessments

EPA Grant Number: R835166
Title: Carolina Center for Computational Toxicology: Assays, models and tools for NextGen safety assessments
Investigators: Rusyn, Ivan , Wright, Fred A. , Yeatts, Karin B. , Tropsha, Alex
Current Investigators: Rusyn, Ivan , Wright, Fred A. , Tropsha, Alexander , Chiu, Weihsueh
Institution: University of North Carolina at Chapel Hill , North Carolina State University
Current Institution: University of North Carolina at Chapel Hill , North Carolina State University , Texas A & M University
EPA Project Officer: Lasat, Mitch
Project Period: July 1, 2012 through June 30, 2016 (Extended to June 30, 2017)
Project Period Covered by this Report: July 1, 2012 through September 30,2013
Project Amount: $1,200,000
RFA: Developing High-Throughput Assays for Predictive Modeling of Reproductive and Developmental Toxicity Modulated Through the Endocrine System or Pertinent Pathways in Humans and Species Relevant to Ecological Risk Assessment (2011) RFA Text |  Recipients Lists
Research Category: Chemical Safety for Sustainability

Objective:

The objective of the Carolina Center for Computational Toxicology is to advance the science and practice of toxicology by (i) filling critical gaps in our knowledge of the toxicity mechanisms, (ii) incorporating the population-based screening methods into the practice of toxicity testing, (iii) developing reliable computational models and tools that address specific existing challenges in hazard identification, and (iv) engaging with the stakeholders to increase the impact of our work. 

Progress Summary:

In Specific objective 1, our goal is to develop a quantitative high-throughput screening (qHTS) approach to probe differential chemical effects in a population-based in vitro system. To this end, we have been following up on the success of our collaboration with NCATS and NIEHS/NTP in which we successfully screened 1,086 human lymphoblast cell lines, representing 9 populations from 5 continents, in a cell viability assay with 179 diverse environmental chemicals at 8 concentrations (Abdo, et al. 2012; Abdo, et al. 2013). A follow-up study (Abdo, et al. 2014) was designed to address two limitations of a lymphoblast-based in vitro screening model: restricted metabolic capacity of lymphoblasts and the potential to screen complex mixtures. We screened two environmental pesticide mixtures (organochlorine pesticide environmental mixture extracted from a passive surface water sampling device, or a mixture of 36 currently used pesticides) and drug/metabolite pairs (Carbamazepine, Sulfamethoxazole, or two major metabolites of each drug). These data provide the opportunity to establish population-based confidence intervals in cytotoxicity, as well as probe candidate susceptibility pathways. In addition, as an alternative way of addressing the limitations of lymphoblast cells with regards to metabolism and target-specific toxicity, we collaborated with Molecular Devices and Cellular Dynamics, companies that are world leaders in high-content/high throughput cellular imaging and induced pluripotent stem cell (iPSC)-based in vitro models, respectively. In these studies (Sirenko, et al. 2013a; Sirenko, et al. 2013b; Sirenko, et al. 2013c), we used iPSC-derived hepatocytes and cardiomyocytes to screen large (100+) libraries of chemicals and determined how the multi-parametric assessment of concentration response toxicity phenotypes can be used to make hazard-based rankings.

In Specific objective 2, our goal is to provide the computational toxicology solutions for risk characterization in NexGen assessments with a focus on point-of-departure and population variability. First, we are developing computational solutions for estimation of the population variability in toxicity by utilizing the power of the 1000 Genomes Toxicity Screening data (Abdo, et al. 2012; Abdo, et al. 2013). Specifically, we partnered with Sage Bionetworks (Seattle, WA) to use this data for one of DREAM (Dialogue for Reverse Engineering Assessments & Methods) Challenges. In sub-challenge 1, the participants were asked to predict inter-individual variability in in vitro cytotoxicity based on genomic profiles of individual cell lines. For each compound, participants were challenged to predict the absolute values and relative ranks of cytotoxicity across a set of unknown cell lines for which genomic data is available. For sub-challenge 2, the task was for each compound, predict the concentration at which median cytotoxicity would occur, as well as inter-individual variation in cytotoxicity, described by the 5-95th percentile range, across the population. Each prediction was scored based on the participant’s ability to predict these two parameters within a set of compounds excluded from the training set. The NIEHS-NCATS-UNC Toxicogenetics Challenge attracted a “crowd” of ~250 researchers who used these data to elucidate the extent to which adverse effects of compounds can be inferred from genomic and/or chemical structure data. There were 99 models submitted by 35 teams for sub-challenge 1, and 85 models by 23 teams for sub-challenge 2. Final announcement of the winners of the challenges will be made at the 2013 DREAM conference that will held on November 8-12 in Toronto, Canada, in conjunction with the RECOMB/ISCB Conference on Regulatory and Systems Genomics. 

Second, we are developing computational solutions for organ-specific toxicity using iPSCs. We used the dose-response data from iPSC-based studies to assess prediction accuracy of the individual parameters, as well as the multi-parametric data that was integrated by means of point-of-departure information into a single prediction using ToxPi software.
 
Third, we are developing computational solutions for estimation of the point-of-departure. In response to the need to develop default approaches to support risk estimation for chemicals lacking chemical-specific information, we are developing the Conditional Toxicity Value (CTV) Predictor in collaboration with EPA/NCEA (Wignall et al. 2013a). CTV is an approach to support risk estimation for chemicals lacking chemical-specific information. CTV tool uses chemical properties and limited experimental data to predict toxicity values (such as the oral slope factor, inhalation unit risk, reference dose and concentration), as well as points of departure. Toxicity potency ranking can also be generated for groups of chemicals. CTV predictions rely on a new comprehensive database of existing toxicity values, the associated points of departure (with benchmark doses calculated where feasible) and other experimental data. The approach combines QSAR and regression modeling, and incorporates OECD principles for model building and external cross-validation. To enable development of CTV for point-of-departure values (e.g., BMD or LOAEL), we applied a standardized process for conducting BMD modeling to reduce inconsistencies in model fitting and selection and to identify study design features affecting BMD modeling fit acceptability (Wignall, et al. 2013b). We evaluated dose-response data (880 datasets) for 352 environmental chemicals with existing human health assessments. We calculated benchmark doses (1 standard deviation or 10% response, BMD10/1SD) for each chemical in a standardized way with pre-specified criteria for model fit acceptance. 
 
Fourth, we are developing computational solutions for cloud-based development of human health assessments of chemicals. To this end, we are developing HAWC (Health Assessment Workspace Collaborative, https://hawcproject.org/), a modular, cloud-ready, informatics-based system to synthesize multiple data and information (Shapiro, et al. 2013). HAWC seamlessly integrates and documents the overall workflow from literature search and review, data extraction and evidence synthesis, dose-response analysis and uncertainty characterization, to creation of customized reports. Crucial benefits of such a system include improved integrity of the data and analysis results, greater transparency, standardization of data presentation, and increased consistency. By including both a web-based workspace for assessment teams who can collaborate on the same assessment rather than share files and edits, and a complementary web-based portal for reviewers and stakeholders, all interested parties have dynamic access to completed and ongoing assessments. 
 
In Specific objective 3, we develop cheminformatics-based, as well as enhanced chemical-biological, models of in vivo reproductive and developmental toxicity that rely on concomitant exploration of chemical descriptors and population-based screening data. In order to develop in silico predictors to identify Estrogen Receptor (ER)-mediated endocrine dedisruption, we collected from public databases and scientific literature a large number of ER ligands along with reported relative binding affinity to ERα and/or ERβ. A novel multi-task learning QSAR modeling approach was applied to develop models capable of predicting the binding affinity of ligands to both ER subtypes. In addition, as a complementary approach, docking studies were performed on a set of ER agonists/antagonists and corresponding presumed decoys/non-binders. Virtual screening of an uterotrophic dataset validated that the consensus of MTL QSAR and docking models had the highest enrichment power. Virtual screening of the EPA Tox21 library yielded a prioritized list of 286 putative estrogenic compounds for future in vitro and in vivo tests on endocrine disruption (Zhang et al. 2013). Similar studies on the thyroid hormone receptor beta are ongoing. 
 
In addition, we developed a new method termed Chemical Biological Read Across (Low et al. 2013), which infers each compound's toxicity from both chemical and biological analogues whose similarities are determined by the Tanimoto coefficient. CBRA-based hazard classification exhibits consistently high external classification accuracy and applicability to diverse chemicals and diverse “biological” datasets. Transparency of the CBRA approach is aided by the use of radial plots that show the relative contribution of analogous chemical and biological neighbors. Identification of both chemical and biological features that give rise to the high accuracy of CBRA-based toxicity prediction facilitates mechanistic interpretation of the models. 

 

Future Activities:

In Specific Objective 1, we will finalize the analysis of the 1000 Genomes Project screening; finalize the analysis of the population-wide experiment with mixtures and drug/metabolite pairs; and further explore the utility of iPSC models for population-based high-content/high-throughput screening by developing additional collaborations with Cellular Dynamics who are establishing iPSCs from hundreds of individuals with sequenced genomes. 

In Specific Objective 2, we will work with the winners of the NIEHS-NCATS-UNC Toxicogenetics Challenge to develop user-friendly and publicly available computational approaches based on the best-performing models; finish development of chemical structure- and biological data-based CTV; finish and deploy HAWC; continue working with U.S. federal agencies and other stakeholders to improve functionalities in HAWC. 
 
In Specific Objective 3, we will work with EPA Office of Research and Development and Office of Chemical Safety and Pollution Prevention on applying ER models to chemical prioritization for in vivo screening; and finish development of QSAR and docking models for THR. 

References:

Abdo N, Marlot P, Pirmohamed M, Shea D, Wright FA, Rusyn I. 2014. Utilizing human population based in vitro model to investigate pesticide mixtures and drug/metabolite pairs. In: Society of Toxicology Annual Meeting. Phoenix, AZ.

Abdo N, Xia M, Brown CC, Kosyk O, Huang R, Sakamuru S, et al. 2015. Population-based in vitro hazard and concentration-response assessment of chemicals: The 1000 genomes high throughput screening study. Environ Health Perspect:(in press).

Grimm FA, Iwata Y, Sirenko O, Crittenden C, Roy T, Boogaard PJ, et al. 2015. Toxicological categorization of petroleum substances through high-content screening of induced pluripotent stem cell (ipsc) derived cardiomyocytes and hepatocytes In: Annual Meeting of the Society of Toxicology. San Diego, CA.

Politi R, Rusyn I, Tropsha A. 2014. Prediction of binding affinity and efficacy of thyroid hormone receptor ligands using qsar and structure-based modeling methods. Toxicol Appl Pharmacol 280:177-189.

Reif DM, Sypa M, Lock EF, Wright FA, Wilson A, Cathey T, et al. 2013. Toxpi gui: An interactive visualization tool for transparent integration of data from diverse sources of evidence. Bioinformatics 29:402-403.

Shapiro AJ, Cook N, Ross PK, Fox J, Cogliano V, Chiu WA, et al. 2013. Web-based benchmark dose modeling module as a prototype component of an informatics-based system for human health assessments of chemicals. In: Society of Toxicology Annual Meeting. San Antonio, TX.

Sirenko O, Crittenden C, Callamaras N, Hesley J, Chen YW, Funes C, et al. 2013a. Multiparameter in vitro assessment of compound effects on cardiomyocyte physiology using ipsc cells. J Biomol Screen 18:39-53.

Sirenko O, Cromwell EF, Crittenden C, Wignall JA, Wright FA, Rusyn I. 2013b. Assessment of beating parameters in human induced pluripotent stem cells enables quantitative in vitro screening for cardiotoxicity. Toxicol Appl Pharmacol 273:500-507.

Sirenko O, Hesley J, Rusyn I, Cromwell EF. 2014. High-content assays for hepatotoxicity using induced pluripotent stem cell-derived cells. Assay Drug Dev Technol 12:43-54.

Sirenko O, Hesley J, Rusyn I, Cromwell EF. 2015. High-content high-throughput assays for characterizing the viability and morphology of human ipsc-derived neuronal cultures. Assay Drug Dev Technol:in press.

Wignall JA, Muratov E, Fourches D, Tropsha A, Woodruff T, Zeise L, et al. 2013. Conditional toxicity value (ctv) predictor for generating toxicity values for data-sparse chemicals. In: Society of Toxicology Annual Meeting. San Antonio, TX.

Wilson MR, Ball N, Carney EW, Rowlands JC, Rusyn I. 2015. Data integration and visualization for transparent communication of the category read across using toxpi (toxicological priority index) tool: P-series glycol ethers case study. In: Annual Meeting of the Society of Toxicology. San Diego, CA.

World Health Organization. 2014. Guidance document on evaluating and expressing uncertainty in hazard characterization. Harmonization document no. 11. Geneva, Switzerland.

Yeakley J, Abdo N, Chappell G, Shepard P, Rusyn I, Seligmann B. 2015. A cost effective targeted sequencing method for monitoring gene expression. In: Annual Meeting of the Society of Toxicology. San Diego, CA.


Journal Articles on this Report : 6 Displayed | Download in RIS Format

Other project views: All 53 publications 39 publications in selected types All 39 journal articles
Type Citation Project Document Sources
Journal Article Low Y, Sedykh A, Fourches D, Golbraikh A, Whelan M, Rusyn I, Tropsha A. Integrative chemical-biological read-across approach for chemical hazard classification. Chemical Research in Toxicology 2013;26(8):1199-1208. R835166 (2013)
R835166 (2016)
R835166 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: ACS-Full Text HTML
    Exit
  • Abstract: ACS-Abstract
    Exit
  • Other: ACS-Full Text PDF
    Exit
  • Journal Article Sirenko O, Cromwell EF, Crittenden C, Wignall JA, Wright FA, Rusyn I. Assessment of beating parameters in human induced pluripotent stem cells enables quantitative in vitro screening for cardiotoxicity. Toxicology and Applied Pharmacology 2013;273(3):500-507. R835166 (2013)
    R835166 (2014)
    R835166 (2016)
    R835166 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Abstract: ScienceDirect-Abstract
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Journal Article Sirenko O, Crittenden C, Callamaras N, Hesley J, Chen YW, Funes C, Rusyn I, Anson B, Cromwell EF. Multiparameter in vitro assessment of compound effects on cardiomyocyte physiology using iPSC cells. Journal of Biomolecular Screening 2013;18(1):39-53. R835166 (2013)
    R835166 (2016)
    R835166 (Final)
  • Abstract from PubMed
  • Full-text: Sage-Full Text HTML
    Exit
  • Abstract: Sage-Abstract
    Exit
  • Other: Sage-Full Text PDF
    Exit
  • Journal Article Sirenko O, Hesley J, Rusyn I, Cromwell EF. High-content assays for hepatotoxicity using induced pluripotent stem cell-derived cells. Assay and Drug Development Technologies 2014;12(1):43-54. R835166 (2013)
    R835166 (2014)
    R835166 (2016)
    R835166 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Abstract: Liebert-Abstract
    Exit
  • Journal Article Wignall JA, Shapiro AJ, Wright FA, Woodruff TJ, Chiu WA, Guyton KZ, Rusyn I. Standardizing benchmark dose calculations to improve science-based decisions in human health assessments. Environmental Health Perspectives 2014;122(5):499-505. R835166 (2013)
    R835166 (2014)
    R835166 (2016)
    R835166 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: EHP-Full Text PDF
  • Abstract: EHP-Abstract and Full Text HTML
  • Journal Article Zhang L, Sedykh A, Tripathi A, Zhu H, Afantitis A, Mouchlis VD, Melagraki G, Rusyn I, Tropsha A. Identification of putative estrogen receptor-mediated endocrine disrupting chemicals using QSAR- and structure-based virtual screening approaches. Toxicology and Applied Pharmacology 2013;272(1):67-76. R835166 (2013)
    R835166 (Final)
    R833825 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Supplemental Keywords:

    Bioinformatics, biostatistics, computational toxicology, QSAR, ToxCast, high throughput screening 

    Progress and Final Reports:

    Original Abstract
  • 2014 Progress Report
  • 2015
  • 2016 Progress Report
  • Final Report