NORA Manufacturing Sector Strategic Goals
927ZHNL - Computational Toxicology for the Manufacturing Sector
Principal Investigator (PI)
Primary Goal Addressed
Secondary Goal Addressed
Attributed to Manufacturing
This project will explore the utility of the computer in assimilating toxicology information relevant to chemical exposures in the manufacturing sector. A chemical exposure of occupational and manufacturing significance will be selected and purported biomarkers associated with that chemical will be mined from publicly available scientific sources. Once collected, a computational toxicology approach will be used to systematically identify, document, validate and incorporate biomarkers into an occupational risk assessment.
The project will determine if sufficient data can be gleaned from the literature and publicly available databases to identify occupational chemical hazards, validate a biomarker for biological monitoring and construct an appropriate OEL. The project is composed of the following steps:
Selection of Manufacturing Chemical Exposure – One of three (1-bromopropane, 1,3-butadiene, diacetyl) occupationally relevant chemicals will be selected as the case-study exposure using these criteria: 1) prevalence in the workplace; 2) severity of toxic response; 3) availability of biomarker data; 4) need for an updated OEL
Biomarker Data Search – Biomarker data will be extracted from the National Library of Medicine TOXNET and Medline databases, from comprehensive reviews, and by contacting researchers familiar with the case-study chemical.
Biomarker Database-Previously we developed a customized a database for diverse biomarker data which included: 1) general study information; 2) species or cell type; 3) specific biomarker information; and 4) dose-response data.
Categorization of Inputs - Studies will be categorized by biomarker types: exposure, internal dose, effective dose, early effects, mild effects, severe effects, or susceptibility. Disease endpoints specific to the case-study chemical will be identified.
Data Evaluation - Biomarker data will be normalized to facilitate inter-study comparisons and combining studies for validation analysis. Common endpoints from different studies are converted to the same response units, and then normalized. Data will be collected from several studies of different groups with different exposure scenarios and experimental designs. To enable quantitative analysis, data will be grouped into exposure categories determined by examining the range of exposure associated with the various candidate biomarker and disease observations. The data will be ordered by increasing exposure concentration, and exposure bands will be chosen.
Biomarker Analysis and Validation – This approach resembles a decision support system that incorporates a suite of analysis validation tools. The tools used for a given chemical will vary depending on the dataset available. With benzene, the analytical tools were qualitative analysis using the Hill criteria, qualitative graphical analysis, linear regression, and Bayesian network modeling. Sufficient data were available to use the Bayesian network to quantitatively incorporate the biomarker data into the dose-response analysis. Other chemical data sets may require only simple regression tools to identify the appropriate biomarkers, or potential mechanisms of action may not be sufficiently known to implement these more powerful quantitative approaches. The various biomarkers will be ordered into hypothesized pathways based on knowledge of mechanisms of action to lead to the relevant endpoints. The first approach for validation and analysis of the pathway(s) is application of the Hill criteria.
Bayesian network - The Bayesian network will be based on the hypothesized mode of action, the availability of adequate data from the screened pool of potential studies, and the results of the Hill criteria-based biomarker evaluation. The Bayesian network is a series of power and logistic regression equations describing the relationships between the nodes in the network. Briefly, each marker or disease endpoint (i.e., each node) is represented by an equation with a constant term, and one slope term for each of the node's parents. The network will be calibrated via Markov Chain Monte Carlo (MCMC) simulation to find distributions for the parameters of the regression equations. This calibration is followed by analyses to choose among competing biomarkers and to identify elements of the network of potential biomarkers that are most predictive of disease.
Dose-Response Analyses - Dose-response analyses for the selected disease endpoint incorporating the validated biomarkers will be conducted using four approaches. First, a logistic model evaluates the relationship between the case-study chemical concentration and the selected disease. This dose-response provides a reference for evaluating the impact of incorporating the biomarkers. Second, a logistic model evaluates the relationship between exposure biomarkers and the disease, in order to derive the effective exposure biomarker value corresponding to the theoretical risk used as the basis for development of OELs (i.e., 1/1000 incidence). The third approach employs the validated precursor immediately preceding the disease as the dependent variable. Finally, a Monte Carlo analysis of the Bayesian network model is conducted using MCSim to calculate the case-study concentration that, when propagated through the biomarker network, produces a disease response rate of 1/1000.
Objectives and time frame: The specific aims of the project are to:
Select an occupationally relevant and manufacturing specific case-study chemical exposure (Q2 FY09)
Populate the biomarker database with biomarkers relevant to the selected case-study chemical exposure (Q4 FY09)
Categorize the biomarker data (Q1 FY10)
Analyze and identify validated biomarkers (Q2 FY10)
Build an exposure-to-disease biomarker pathway for the validated biomarkers (Q3 FY10)
Develop a biomarker-based occupational exposure limit (Q4 FY10)
Select a new occupationally relevant and manufacturing specific case-study chemical exposure and repeat the process. (Q1 FY11)
The primary objective of this project will be the development of a biomarker-based occupational exposure limit (OEL) for a chemical of significance to the manufacturing sector. It is proposed that this OEL will have been developed from publicly available databases and other sources of research information using computer-facilitated data mining and analysis techniques. This non-laboratory approach to developing OEL will be cost effective and efficient, an asset in difficult research resource periods. The project will be evaluated by acceptance of the approach in appropriate scientific journals (Regulatory Toxicology), the number of citations accumulated, interest and invitations to present at medical centers and schools of public health. Ultimately, the evaluation will be determined by application to the manufacturing sector in the form of accepted OELs for manufacturing chemicals and/or the use of validated biomarkers for medical surveillance, intervention and biomonitoring.
Exposures to toxicants in the manufacturing work setting contribute significantly to the country's disease burden. Identifying, preventing, reducing and ameliorating risks from these exposures is a critical public health need. While there are 750 occupational exposure limits (OELs) in practice, 2/3's are outdated and require revisions, and an estimated 3000 chemicals require new OELs. The cost of necessary investigations regarding these exposures and the expertise required to evaluate the information and convert it to OELs is prohibitive. Computational toxicology offers a state-of-the-art, cost efficient approach to generating important answers to complex occupational safety and health problems.
Manufacturing Sector Strategic Goal 6: Reduce the incidence and prevalence of cancer due to exposures in the manufacturing sector.