SAPPHIRE’s flexibility showcases an important lesson about Semantic Web systems: once they are configured for a general problem—in this case, public health reporting—they can quickly be adapted to a variety of situations within that field. Indeed, the CDC would like to roll out a single, integrated, SAPPHIRE-style illness alert system nationwide.
SAPPHIRE succeeds because it can unify information from many places, which can then be used for different goals. This same attribute is fueling FOAF’s grassroots growth. By using an agreed-on Semantic Web vocabulary, the FOAF system finds common interests among friends and acquaintances, even if they do not belong to the same social-networking sites such as MySpace or Facebook. FOAF enthusiasts are also now developing semantic trust networks—white lists of trusted senders—as a way to fight e-mail spam.
Crossing Boundaries The success of SAPPHIRE and other applications has prompted calls for more Semantic Web integration in health care. The Food and Drug Administration and the National Institutes of Health have both recently declared that a shift toward research into translating data across boundaries is necessary for improving the drug development and delivery process.
The same work will enhance the traditional computerized clinical decision support (CDS) systems that medical professionals use—knowledge bases that contain the latest wisdom on therapeutic treatments. Each hospital, physicians’ network and insurance company has had to custom-design its own system, and all of them are struggling mightily to stay current. Every time an advance is made about diagnoses, clinical procedures or drug safety—which is often—administrators must rework their systems. The personnel time required is usually far greater than most of these organizations can afford. Furthermore, because the custom systems are frequently incompatible, making industry-wide insights or deciphering best practices is slow and cumbersome. What is more, “we are investigating Semantic Web technologies because traditional approaches for data integration, knowledge management and decision support will not scale to what is needed for personalized medicine,” says John Glaser, chief information officer at Partners HealthCare System in Boston.
To remedy this situation, Agfa HealthCare has constructed a prototype CDS system based on Semantic Web technologies. When a person inputs a change into one part of a system, rec-ords that should be altered in other parts of the system or in the systems of another institution are automatically updated. For example, Agfa’s prototype transforms standard radiology protocols into Semantic Web notation and integrates them with other common knowledge bases, such as clinical guidelines from medical societies. Institutions can maintain their own internally standardized content, yet end users such as hospitals can readily integrate new content, greatly reducing the labor hours required.
As systems such as Agfa’s are implemented across the health care network, medical knowledge bases will become smarter, easier and less expensive to use. Imagine a patient who is prone to blood clots and has a genetic mutation that, according to current medical literature, will respond well to a new anticlotting medication. Over the ensuing months, however, new studies show that particular variants of this mutation actually cause that same drug to increase clotting. This patient’s clinician must be notified to change the therapy for anyone with this variant. How could notifications such as this be effectively carried out given that thousands of genes are involved in hundreds of diseases across millions of patients? Meeting this challenge will not be possible without robust semantic approaches.
Daily Life, Too The same Semantic Web technologies that are transforming drug discovery and health care are being applied to more general situations. One example is Science Commons, which helps researchers openly post data on the Web. The nonprofit organization provides Semantic Web tools for attaching legally binding copyright and licensing information to those data. This capability allows a scientist, for example, to instruct a software applet to go find information about a particular gene—but only information that comes with a free license.
DBpedia is an effort to smartly link information within Wikipedia’s seven million articles. This project will allow Web surfers to perform detailed searches of Wikipedia’s content that are impossible today, such as, “Find me all the films nominated for a Best Picture Academy Award before 1990 that ran longer than three hours.”
As applications develop, they will dovetail with research at the Web consortium and elsewhere aimed at fulfilling the Semantic Web vision. Reaching agreement on standards can be slow, and some skeptics wonder if a big company could overtake this work by promoting a set of proprietary semantic protocols and browsers. Perhaps. But note that numerous companies and universities are involved in the consortium’s semantic working groups. They realize that if these groups can devise a few well-designed protocols that support the broadest Semantic Web possible, there will be more room in the future for any company to make money from it.