June 13, 2007
The Art of Visualization
Visualization expert George Grinstein, director, Center for Biomolecular & Medical Informatics, University of Massachusetts, and Peter Henstock, Pfizer Research Technology Center spoke of the importance of their craft.
Properly constructed and used, said Grinstein, visualization not only makes understanding complex data more accessible, but also can become a tool for data exploration. He showed several approaches to displaying multi-dimension data that didn’t confound the eye or mind including a software tool that was able to de-convolute a confusing protein network into an orderly and visually intuitive pathway diagram. Pfizer’s Henstock is collaborating with Grinstein on several visualization projects. Most often, said Henstock, those engagements are around specific projects, but the resulting techniques also often make their way into standard tools for use by Pfizer researchers.
Talapady Narayana Bhat, project leader bioinformatics, National Institute of Standards and Technology, gave an update on efforts by NIST and NCI to create a structure library using visual icons and semantic web technology. He showed work on an HIV Protease Database Gallery.
Shaping a Portfolio
According to Pfizer’s David de Graaf, systems biology is now part of pharma’s core business. Integrating information from pharmacodynamics and pharmacokinetics can shorten time to market. Meanwhile, modeling helps people to abstract on a number of levels, from cell signaling to cell cycles, and provide links to extrapolate between these levels using ‘omic data sets.
For example, de Graaf noted how Pfizer is interested in p38 inhibitors and oral analogues of TNF regulators. Could it optimize investment decisions by looking at cell signaling, cell behavior, cell physiology, and/or mechanistic models? One route is to use text-mining software. De Graaf uses software from Linguamatics, Cognia, Teranode, and QUOSA to move to a broader understanding before making major investments in specific compounds. He estimated that $18 billion is spent per year on compounds that never reach market, while $30 billion is spent reinventing what is in the literature.
Biogen-Idec’s William Hayes estimated that 1 in 4 projects undergo attrition for reasons already documented in a body of literature costing roughly $1 trillion in the past 15 years. Drug safety and toxicology issues clearly call for heavy-duty literature informatics, and can justify both financial and time investments in a big way, he said.
Many query examples Hayes described were not possible until recently — from identifying interacting proteins, interesting biological co-occurrences and author co-occurrences (supernodes) to finding in-licensing opportunities or competitive intelligence from text mining of patents. Biogen-Idec uses a variety of software, lots of storage, and recognizes that this is bleeding edge and still under development.
Building a Bigger Boat
Lisa Gerrard (Johnson & Johnson) calls translational medicine the interface between discovery and clinical research, and says the field needs LIMS, workflow tools, and analysis tools to succeed. She called for tool development highlighted by tight integration and co-development with scientists, flexibility, and user-friendly formats.
Don Rule (Microsoft) sees the need for better tools as well. Applying Moore’s law to life sciences, he said “IT-centric views” are the only way to accelerate solutions to keep up with the growing data.
XML, he argued, will improve scientific interoperability across workgroups and companies. His view of personalized medicine, fueled by biomarker advancement, requires a shift in the development paradigm. Rather that pushing forward from drug discovery through development and clinical evaluation to commercialization, Rule promised that PK/PD modeling and post-market surveillance would connect the steps in a growing feedback loop.
Genomics, gene expression, and imaging are all poised for a data explosion, and IT has to meet those needs. Recalling the line from Jaws, Rule promised, “We’re going to need a bigger boat.”
Oracle’s Life Science User Group
At the third annual Life Science User Group, organizer Charles Berger, senior director of product management, said he was delighted with all 22 presentations showcasing Oracle’s database technology. “Whether it was simple data storage of huge volumes of data, advanced statistical function and data mining, or implementing semantic web technologies, there was something for everyone,” he said.
John Quackenbush, professor of computational biology and bioinformatics, Dana-Farber Cancer Institute, argued that genomics has transformed biomedical research through the technologies it has spawned. “Genomic technologies are turning clinical and laboratory sciences into information sciences,” says Quackenbush. “There are two ways to beat the data overload problem that genomic technologies engender: even more data and effective data management and information integration strategies.”
Quackenbush’s group has an Oracle Commitment Grant to begin developing an integrated warehouse-style solution uniting clinical, research, and public domain data. An ongoing challenge is protecting patient information and confidentiality without completely stripping it of value. “Our initial focus is on moving information from the clinic to the lab, but we cannot neglect the problem of going from the lab back to the clinic. We have to recognize that a database is a model built on how people use data to address questions,” Quackenbush says.
— Compiled by John Russell, Mary Chitty, Allison Proffitt, Kevin Davies
Subscribe to Bio-IT World magazine.