YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

Intelligent Drug Surveillance

Best Practices Winner: AstraZeneca
Project: The Safety Intelligence Program
Category: Editors’ Choice
Nominated by: BioWisdom

By Alissa Poh

July 20, 2009
| Good collaborations in drug discovery are often “happy coincidences,” says David Cook, AstraZeneca’s associate director for global safety assessment, when describing the company’s partnership with BioWisdom in developing the Safety Intelligence Program (SIP).

“Back in 2007, AstraZeneca was focusing on how best to affect early decisions in drug discovery; how we could influence the daily decisions our project chemists and pharmacologists were making, and get them thinking about safety assessment. Julie [Barnes] rang me up around then and said BioWisdom was thinking of formulating a general problem-solving approach to toxicology issues in drug discovery. She asked if we’d like to come on board.”

The result was SIP, described as the “largest forever-expanding collection of known chemical effects occurring in different tissues, drug effects on clinical biomarkers of tissue injury, and drug molecular mechanisms.” Currently, SIP contains almost 100,000 individual facts, or “assertions,” related to the liver’s response to more than 5,500 different compounds in over 20 species.

The collaborators began with the liver, as acute hepatic injury is among the most common forms of preclinical and clinical toxicities seen with drugs, responsible for more than 30 percent of all drug withdrawals.

“The word ‘idiosyncratic’ is key in liver injury,” says Julie Barnes, BioWisdom’s chief scientific officer. “Those unexpected, adverse reactions in humans that we just can’t predict, because we don’t know enough of a given compound’s biology, in the context of individual patients.”

“Our project teams always want to do the best job possible, but the mechanisms of retrieving and analyzing information they need are often tedious,” Cook says. “I was thus very intrigued by the application of text logic and mining to the problem of retrieving data of interest from the pile.”

Meanwhile, BioWisdom had begun dealing with the problem of language inconsistencies in medical literature and regulatory documents by building Sofia, an ontology-based platform capable of liberating intelligence from multiple source formats. The coalescence between both companies, then, was perfectly natural, with the idea of exploiting this platform as a foundation for SIP. Sofia was used to generate liver-related “assertional metadata,” which comprises thousands of highly accurate and comprehensive key observations distilled from over 19 million documents and database records.

A key feature of SIP is that these assertions are all rendered in a semantically consistent format, with an accuracy level of 97 percent. “There’re so many ways someone will describe a disease, protein or drug—for example, problems associated with bile excretion can be called biliary stasis, cholestasia, or cholestatic injury,” says Jane Reed, a principle consultant in BioWisdom’s health care sector and team leader for this particular project. “So if you’re trying to work out a particular compound’s side effects, you need to know the different descriptions, which our vocabularies cover. Users can also clearly visualize such diverse information, as our technology pulls it all together.” 

Verb Relationships
In other words, SIP very much highlights the science of ontology. “We use the language of verb forms and their relationship with nouns to sort out what things really are,” Barnes explains, “so in defining an obscure statement like ‘AZ binds BCEP,’ the verb ‘binds’ immediately implicates AZ as the chemical Aztreonam and not AstraZeneca, because companies don’t bind proteins!”

SIP’s nascent form was delivered in January 2008. Its potential was evident even then, Reed says, but it wasn’t applied to AstraZeneca’s business issues until version 2.0—complete with in-depth liver data—was released in September. The program’s improved user functionality helped drive the collaborators’ decision to submit SIP for this year’s Best Practices competition in the Drug Discovery and Development category.

“Really, it just felt like the right time,” Reed says. “But at the awards ceremony, when they announced the winners [Amgen and Genedata] in our category, everyone in our group was disappointed. Then they announced the Editor’s Choice award at the end—we were taken by surprise and thrilled.” 

The collaborators have extended SIP’s focus to the cardiac system. Next, they’ll hone in on kidney issues. Besides such detailed tissue-specific mining as goes into the development of SIP data, Cook says, the program also provides a sweeping glance at all other toxicities within its vocabulary.

“We’re looking to use similar approaches to leverage our internal knowledge,” he adds. “There’s a huge amount of information within AstraZeneca on compounds that went through rigorous testing but never got out into the clinic—we’d apply the same exercise here, pulling out relationships and generating assertions, to learn from our home history.”

Moreover, says Barnes, “we’ve been discussing drug signatures and how to develop patterns from our data that take us all the way into later phases of the clinical pharmacovigilance space.” The opposite end, in effect, of AstraZeneca’s business needs in early risk detection, and an expansion of the safety arena.

“It’s about breaking down the clinical barrier both ways,” Cook says. “How do we translate both forward to clinical situations, then take what we learn and recycle it to the earlier stages of drug discovery, to make sure we’re really using that information to influence the next generation of drug development?”

Ultimately, it’s about generating new knowledge from old, using SIP analyses to form new hypotheses that can then be explored with actual experimentation. “That, I think, is really exciting stuff, because toxicology is such a broad area to work in,” Cook says. “To focus on a few key areas at a time minimizes our chances of charging up blind alleys.”

This article also appeared in the July-August 2009 issue of Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply today.

Click here to login and leave a comment.  


Add Comment

Text Only 2000 character limit

Page 1 of 1

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.