Semantic technology is gradually finding its way into a wide range of life science applications but has a way to go before it moves beyond its current base of expert users, according to speakers and attendees at last week's Conference on Semantics in Healthcare and Life Sciences.
At the conference, held in Boston, advocates of the technology discussed current efforts to apply semantics in clinical settings and within drug discovery pipelines and also identified some areas that could do with improvement.
Several presentations focused on projects that use semantics to mine information stored in electronic health records or personal health records for use in clinics while others focused on methods for integrating data on chemical compounds, genes, pathways, and drug interactions to identify drug targets and repurpose drugs.
On the health IT front, Isaac Kohane, director of the informatics program at Children's Hospital Boston and co-director of Harvard Medical School's Center for Biomedical Informatics, presented the Substitutable Medical Apps Reusable Technologies, or SMART, system, which aims to enable users to substitute components from different health IT systems without breaking the entire workflow.
In a paper published in the New England Journal of Medicine in 2009, Kohane and colleagues explained that substitutable apps require a system that is "sufficiently modular and interoperable so that a primary care provider could readily use a billing system from one vendor, a prescription-writing program from another, and a laboratory information system from yet another."
Furthermore, "just as consumers may swap out applications on their iPhones, physicians should be able to readily replace one referral-management system with another," the researchers explained in the paper.
The current system consists of a common application programming interface through which the apps interact with the health data; software that helps developers build applications; and SMART-enabled "containers" that sit on EMRs and personal health records and allow the apps to move from one system to the next, Kohane explained at the conference.
On the drug development side, Nadia Anwar, who works for UK-based General Bioinformatics, described her group's efforts to identify drug targets using a linked data approach to combine information from different experiments. The team used the method to determine the effect of small molecules on biological systems and genetic interaction networks in order to pare down a list of drug targets for an agrochemical company.
James Snowden, senior informatics scientist at biopharmaceutical firm UCB Celltech, presented two linked data systems — the Target Information Platform and the Disease Information Platform — that the company developed to gather information about targets and diseases and make it available in a single portal for researchers.
Another semantic drug discovery tool was developed by researchers at Indiana University Bloomington. Dubbed Chem2Bio2RDF, it semantically links data on chemical compounds, protein targets, genes, metabolic pathways, diseases, and side effects so that researchers can study interactions between drugs and biological systems.
Moving into the post-marketing setting, Richard Boyce, a professor at the University of Pittsburgh, discussed efforts to create a resource to provide updated safety information for drug package inserts.
The tool uses natural language processing to mine information from the scientific literature on drug-drug interactions, age-related changes in clearance, metabolic clearance pathways, and genetic polymorphisms that could affect how drugs are metabolized.
The development team, which also includes researchers at Elsevier and the European Bioinformatics Institute, are focusing on providing updated content for 25 psychotropic medications and have created a prototype of the system for two of these medicines — the antidepressants citalopram and venlafaxine.
Joanne Luciano, an associate professor at Rensselaer Polytechnic Institute and co-chair of this year's CHALS program, told BioInform via e-mail that the technological aspects of the field are "pretty well developed" at present.
However, she acknowledged that the growth of the field will be limited until there are more applications available for non-expert users.
One example could be to create semantic applications that provide individuals with information that enables them to take more responsibility for their health and "fight the marketing bombardment that creates their poor eating and lifestyle habits," she said
Another challenge Luciano cited is "getting meaningful data into the system" — a point that was echoed by other attendees throughout the conference.
For example, Ron Calvanio, a professor of neurology at Harvard Medical School, highlighted in his presentation how a dearth of data poses problems in providing care for some patients.
Calvanio explained in his presentation that often the data needed to analyze complex neurological disorders — such as contextual information like the time of day when a particular symptom is observed — isn't stored in patient medical records.
He has attempted to capture this missing information using a software tool he developed called SymTrend, which patients can use to collect data on their symptoms and routine behaviors that trigger these symptoms. This data has to be analyzed visually in order to identify useful patterns that could be used to improve the individual's care and overall health — a rather time-consuming task, Calvanio said.
An area where semantic technologies could be useful in this scenario would be in automatically recognizing patterns in the data collected by the patients and then linking that information to literature on mental disorders, he said.
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com.