A review of Data Fusion Methodology and Applications in the context of Dose Response Assessment and Human Health Risk Assessment
Asish Mohapatra, Health Canada, Canada
Rehan Sadiq, Amin Zargar, M. Shafiqul Islam, and Roberta Dyck, University of British Columbia, Canada
In light of National Academy of Sciences (2009) Science and Decisions (The Silver Book) recommendations to integrate data across environmental and toxicological information domain related to various biological organizational levels (e.g., biochemical, biomedical, molecular,-omics, cellular, tissue and ogans), our project evaluated data fusion techniques and methodologies to address the integration of heterogeneous datasets. Based on our review of various data fusion extensions, techniques, and existing conceptual frameworks (e.g., Joint Director Laboratories, JDL-DF framework), we developed a modified data fusion based human health risk analysis (DF-HHRA) framework. Such a framework is capable of integrating multiple sources of information (e.g., dose-response, QSAR predictive toxicological datasets, and knowledge from literature sources and expert judgement) for different components of HHRA. Furthermore, we identified several DF extension and toxicological database management techniques and incorporated them within the proposed DF-HHRA framework. Furthermore, we also developed a system biology framework within our dafa fusion model to integrate data across biological organizational levels. We selected two petroleum hydrocarbon group of chemicals (e.g., Benzene and F1 hydrocarbon mixture) for our data fusion project. Based on our modified JDL DF-HHRA framework, we first identified relevant datasets and databases where we extracted petroleum hydrocarbon toxicology information and then by using extended data fusion architectures and statistical data fusion extension techniques, we integrated spatially distributed exposure data in terms of probability distribution. The collected dose-response data from literature were fused using Dempster-Shafer Theory (DST) p boxes. Then, by fusing exposure and dose-response data, risk was characterized in fused p boxes.
Based on our work completed so far, Our methodological framework for DF-HHRA can pro-actively assess and evaluate various public health issues by a front-end toxicological dataset integration via a dynamic toxicological knowledgebase analysis. These analyses can recognize patterns in toxicological datasets and exposure to contaminants both on a spatial and pave the way for analyzing more complex temporal datasets. As such, it provides a framework for efficient and effective integration of "toxicity pathways" and biological based information into human health risk analysis projects. Based on our future proposed project plans, these data fusion based HHRAs should be able to influence risk management decisions.
This has been a collaborative work between Health Canada Alberta and Northern Region and University of British Columbia researchers. We have presented part of this project in various meeting and proceedings in across North America. Please note, the primary author of this abstract has evaluated some of the emerging informatics tools and platforms (e.g., semantic web and data fusion applications in open source platforms and networks in Life Sciences domain) which can be used to design a true collaborative platform to process toxicological and exposure datasets under a human health risk analysis framework. In 2008, 2009, the primary author of this abstract presented a conceptual semantic web framework at the Society for Risk Analysis conference. The conceptual framework was designed as a Semantic Web Informatics Facilitated Tool (SWIFT) for Dynamic Analysis of Risk Tools (DART) - A SWIFT-DART framework. This will be further explored in future phases of our project. Ongoing and future discussions with Collaborative Toxicological initiatives such as OpenTox may pave the way to initiate such an applied project for practical applications in the areas of application of system biology in assessing chemical classifications and toxicology, in silico toxicology, adverse events analysis, drug discovery and toxicity analysis at different biological organizational levels.
Currently, we are in the process of finalizing several components our project and publishing our work in peer-reviewed journals. Based on this abstract, the primary author (corresponding author) of this abstract/paper, will present a poster paper and a oral presentation at the OpenTox 2011 workshop.
