Nov. 13, 2007 | Four years ago, Dimitris Agrafiotis’ team from recently acquired 3-Dimensional Pharmaceuticals (3DP) was entrusted with building a major global informatics platform, dubbed ABCD (Advanced Biological and Chemical Discovery) for Johnson & Johnson Pharmaceutical Research & Development (JJPRD). More than many big pharmas, J&J scientists were struggling because of the organization’s decentralized structure, and wasting time shuffling data between Excel spreadsheets and HTML tables.
This month, Agrafiotis, VP Informatics at JJPRD, and 21 colleagues publish a detailed paper in the Journal of Chemical Information and Modeling that provides the framework for the construction and applications of the ABCD database*. Currently, nearly 1,200 scientists and executives across J&J’s three major research and early development (RED) sites are using ABCD. (Agrafiotis quickly checks the precise number — 1,182.) The primary data in ABCD hails from the three RED units and their satellites — La Jolla, Calif., Spring House, Penn., and Beerse, Belgium.
Agrafiotis says ABCD has changed significantly in the past few years (see How to Spell Discovery, Bio•IT World, June 2004), even since the scientific paper was submitted. “We’ve added in vivo data, we’ve cleaned up and harmonized a lot of the result types, and added more data feeds such as compound availability. Most importantly, we’ve restructured the fact tables to support more effective multi-dimensional aggregation of data at the compound and biological result levels, in order to make reports a lot more effective at all levels of resolution,” says Agrafiotis.
He sounds insulted when asked if there is any commercial software integrated into ABCD. “Zero,” he says. “It’s all home-grown. The only commercial systems we use are Oracle on the back end and Informatica for the ETL (extraction, transformation, and loading). And a couple of component libraries for the application.”
Management, he says, has been surprisingly open and “extremely supportive” of the ABCD initiative and particularly of doing it in house. “It’s unheard of in most other companies,” says Agrafiotis, “where they either stitch together commercial packages or contract out most, if not all, of the development.”
Particularly impressed with ABCD is John Reynders, the recently appointed CIO for Johnson & Johnson's R&D program. "From the outside in, I was very impressed with the platform," says Reynders. "Now that I'm on the inside, I'm even more impressed with the platform. It's something, frankly, I view as us being able to leverage more broadly across the pipeline."
The scientists seem content too. “By bringing together and integrating data sources from around the world, ABCD has fundamentally changed the way scientists from many disciplines work together,” says Peter Connolly, research fellow in medicinal chemistry at JJPRD. “Think about how the Internet has changed the world during the past decade. That’s what ABCD is doing for collaborative science at J&J — bringing information quickly to the desktop using an intuitive, easy-to-use interface.”
Educated in his native Greece and London, Agrafiotis did a post-doc with Harvard Nobel laureate E. J. Corey before moving to the pharma industry. In 1994, he joined 3DP, building computational tools for combinatorial chemistry and structure-based drug design. That company was acquired by JJPRD in 2003.
ABCD consists of two major components — a data warehouse on the backend, and the “Swiss army-knife” Third Dimension Explorer application (3DX) on the front end. There’s also a workspace web portal for delivering and managing ABCD information. “The warehouse did not exist before we started this initiative,” says Agrafiotis. “3DX was based on a lot of code developed at 3DP over the past ten years. ABCD as a goal was in management’s minds when 3DP was acquired, but the implementation had not been decided.”
Agrafiotis points out that the switch to ABCD has been completely voluntary. Management did not twist anybody’s arm to use the new system. Most of the legacy systems are still in place and operating, but their usage is very light.
Most of the users fall into three groups: computational scientists, project champions and data managers, and technology-savvy scientists. Today, scientists conducting chemical queries routinely go through ABCD, using the query wizard embedded in 3DX. This allows them to directly extract data from the database. Some might also use tools from Scitegic (such as Pipeline Pilot) to manipulate the data. “The query interface allows them to combine chemical and biological queries, search by site, by chemist, by project — they can do it all through 3DX,” says Agrafiotis.
He adds that in Beerse, Belgium, JJPRD scientists are using Project Datasets — pre-packaged queries assembled from ABCD and updated on a daily basis. “These ideas are now permeating across the organization,” he says. The database is primarily used in drug discovery research, but its use is expanding into other functional areas. For example, Centocor, one of J&J’s biotech subsidiaries, is using ABCD for clinical data mining purposes.
Agrafiotis manages an RED IT staff of about 36 people, with a dozen or more people working on various projects under the ABCD umbrella. The team manages a support forum for all users and developers. If anyone has a problem or issue, they can post a message to the forum. ABCD staff receives tens of messages each day.
There is also a scientific advisory team composed of “influential internal scientists who have a passion for informatics,” says Agrafiotis. The IT savvy representatives represent all the sites, and meet monthly — more often if there is an urgent issue. Typical issues the advisory team deals with might include: How to score assay results so that users can gauge whether a compound is good, bad, or indifferent? How to normalize data? How to classify in vivo protocols?
Although hardly impartial, Agrafiotis is convinced that the resource affords J&J a competitive advantage. “Absolutely, no doubt about it,” he exclaims. “Otherwise management wouldn’t support this effort so strongly. Support hasn’t wavered at all, in fact it has intensified.” Connolly agrees: “To my knowledge, there’s nothing in the pharma industry to equal its scope and usability,” he says.
"Although it's had a cheminformatics focus in the past, what I've been very delighted to discover is the architecture is very general," says Reynders. "It's built on SOA and is a multi-tiered environment —it is very easy to add plug-ins and capabilities."
Other companies are grappling with the same kinds of issues, says Agrafiotis. “But frankly, it takes a unique ability to do this themselves, the ability to assemble a disciplined and effective team for this type of initiative to succeed.” Everyone in his team “understands we need coherence and consistency in the way the interfaces are built, in the way the system performs. It’s like the old Lexus tagline: ‘The relentless pursuit of perfection.’”
Agrafiotis cites an example. He was working with a modeler in La Jolla on developing a chemical cartridge prototype. “I wanted to search the entire database in less than a second,” he recalls. “I had the prototype down to a couple of seconds, and it beat the heck out of everything out there. But we didn’t settle for that, we wanted to be able to do it in sub-second timeframes.”
A few days later, it was done, and Agrafiotis says the rest of his team is wired the same way. “The team on ABCD has an obsessive drive for excellence in every single aspect of the project: the application, the database, the overall user experience. We look at every piece of code as a statement about our capabilities.”
The recent hiring of Reynders “will be very positive,” predicts Agrafiotis. Adds Reynders: "I'm looking forward to working with Dimitris to extend it into other areas of informatics — bioinformatics, translational informatics, medical informatics, so it can become a platform that extends across the pipeline."
Not one to rest on his laurels, Agrafiotis is busy making plans for the next phase of the database. “The first release was about downloading data, visualizing, mining, analyzing it,” he says. “ABCD is now well into its second phase — the reengineering of all transactional systems — which will allow the biologists to register data into the database.”
The existing systems, he says, leave much to be desired, as they use old technology, differ from site to site, and use different interfaces: “We’re doing this with the same kind of philosophy and with 3DX as the primary user interface. The same application we use for data analysis and visualization, we also use to fit IC50 curves and upload the results into the database, to manage plates, to register and track reactions performed by the chemists so you don’t need a paper notebook.”
In short, “a coherent, unifying solution.”
Subscribe to Bio-IT World magazine.