srvc-EHSR: Sysrev Version Control - Environmental and Health Related Systematic Reviews - While document reviews (DRs) are not the only requirement of environmental and health-related risk assessments (EHRRAs), this component contributes considerable time and money as each DR costs >$140,000. In the pharma industry alone, average yearly DR expenditure is $16.7M. Currently, most EHRRAs rely on manual curation of data including selection of target documents & parsing key data fields in sources. This process, while integral to accuracy & completeness of EHRRAs, requires hundreds of research hours. While increasing numbers of platforms utilize, or permit utility of, machine learning (ML) models, their complex codebase makes it too cumbersome to deploy new features or integrate ML tools for specific use cases. As a result, risk assessments continue to rely on long, manual document reviews which, in turn, delays accessibility to new data for informed chemical & product safety decisions. Therefore, academia, industry, & government regulators would benefit from a platform leveraging ML to optimize EHRRA related DRs. Insilica, LLC will solve these market challenges via a cloud-based or on-site hosted, application, which is remote-capable and enables EHRRA researchers to build customizable workflows based on standardized, interchangeable DR & ML components. Using the intuitive srvc-EHSR interface, researchers will use natural language inputs to define inclusion/exclusion parameters for an environmental chemical review and required data summary statistics. A list of publications or documents will be returned, along with summarized data metrics for defined fields and research questions across all sources. A flexible algorithm structure will enable Insilica’s proprietary ML tools to be leveraged for reviews, as well as customizable open-source tools be shared across researchers and organizations. Once a custom defined DR is complete, individual sub-components of the review can easily be repurposed for future DRs or shared with new researchers to accelerate longitudinal open-source knowledge aggregation in targeted chemical and environmental human health hazard applications. This project will build on a successful Phase I, in which 1) modular data architecture defined, captured, & synchronized data across varying form factors of literature & documents, 2) front end query coupled with ML tools enabled automatic selection of document sources & data summaries with accuracy similar to expert researchers, and 3) a usability study reported high researcher satisfaction. This Phase II will expand the foundation to a commercially viable product with fully validated software, ML tools, & support infrastructure. First, architecture will be enhanced by cross platform interoperability, mobile responsiveness, and self-hosting. Next, ML accuracy will be improved, and open-source options integrated. User front end tools will also be expanded with onboarding, collaboration, administrator, & reporting options. After components are integrated at the system level, migrated to production server, and tested to compliance standards, it will be deployed in a field study to document technical performance, efficiency, & cost savings.
