Managing the Deluge of Scientific Data

Provenance information in eScience is metadata that's critical to effectively manage the exponentially increasing volumes of scientific data from industrial-scale experiment protocols. Semantic provenance, based on domain-specific provenance ontologies, lets software applications unambiguously interpret data in the correct context. The semantic provenance framework for eScience data comprises expressive provenance information and domain-specific provenance ontologies and applies this information to data management. The authors' " two degrees of separation " approach advocates the creation of high-quality provenance information using specialized services. In contrast to workflow engines generating provenance information as a core functionality, the specialized provenance services are integrated into a scientific workflow on demand. This article describes an implementation of the semantic provenance framework for glycoproteomics. e Science, also known as cyber­ infrastructure, represents a par­ adigm shift in scientific research that lets scientists harness Web­based computing and data resources to achieve their objectives faster, more efficiently, and on an industrial scale. Using remote software and experi­ mental equipment, scientists can not only access but also generate and pro­ cess data from distributed sources. The resulting data deluge demands computing solutions that can use high­quality metadata — specifically, domain­specific provenance infor­ mation — to automatically interpret, integrate, and process data. Such so­ lutions bring real value to scientists by answering domain­specific queries effectively to support knowledge dis­ covery over large volumes of scientific data. But creating provenance infor­ mation of the requisite quality in the heterogeneous, distributed, and high­ throughput environment of eScience is a daunting challenge. We argue that incorporating domain knowledge and ontological underpin­ ning in provenance using expressive domain­specific provenance ontologies is an approach equal to the challenge. This semantic provenance imposes a JULY/AUGUST 2008 47 Semantic Provenance for eScience formally defined domain­specific conceptual view on scientific data (domain semantics), mitigates or eliminates terminological hetero­ geneity, and enables the use of reasoning tools for knowledge discovery. Furthermore, we de­ fine a " two degrees of separation " approach for creating semantic provenance using special­ ized software tools. Unlike many prevalent workflow­engine­centric approaches, these tools refer to domain­specific provenance on­ tologies to create provenance information and are integrated into a scientific workflow on demand. We combine the essential aspects of high­ quality provenance — characteristics, a repre­ sentation model, the creation process, and usage — into a single semantic provenance framework. This framework will pave the way for software agents to interpret experimental data unam­ biguously for effective …