In its search for new therapies for cardiovascular disease, Etexx Biopharmaceuticals plans to wield a mighty weapon: Medline. The 35-year old database of biomedical literature currently holds more than 4,600 journals and a total of 12.7 million records, with an additional 500,000 abstracts added every year. According to Etexx CEO Dede Willis, there’s a wealth of information in the venerable resource that has gone untapped because “there’s so much information available, that it’s a question of being able to assimilate it all.”
The company is planning to end this dry spell with a text-mining software package it acquired via an exclusive license from Harold “Skip” Garner’s lab at the University of Texas Southwestern. The software, called IRIDESCENT (Implicit Relationship Identification by Software Construction of an Entity-based Network from Text), differs a bit from other literature-mining methods. Rather than just focusing on correlations between terms, IRIDESCENT finds implicit relationships between seemingly unrelated items and ranks them by potential relevance. Garner said the software relies on a specially designed biomedical ontology that includes genes, drugs, diseases, and all of their possible synonyms and variant spellings. In addition, he said, an underlying statistical package assesses the relevancy and reliability of all the possible connections that are formed between items.
The killer app for this kind of capability, according to Etexx’s founders, is sniffing out novel indications for known drugs.
Etexx is planning on using IRIDESCENT to weed through innumerable papers describing applications for compounds that have already been proven or disproven, to find links to new uses that have yet to be discovered. “Say compound X works or is efficacious against disease Y. What IRIDESCENT is able to do is read through all of these things and find indirect or implicit relationships between, say, compound X and disease Z, that no one has ever tested,” Willis said.
This search for novel applications of known compounds is inspired by the fact that several blockbuster drugs “were not originally developed for their current use,” Willis said. Indeed, Pfizer’s erectile dysfunction drug Viagra was originally developed for angina, and Pharmacia and Upjohn’s hair-loss remedy Rogaine was intended for high-blood-pressure. While these success stories were the result of trial and error, Willis pointed out that a high-throughput approach to making unexpected connections between compounds and indications should be of great interest to pharmaceutical firms.
A discussion of this approach and its application to cardiac hypertrophy appears in the February 12 advance online issue of Bioinformatics [2004 20(3): 389-398]. Garner and his colleagues mined 4,092 Medline articles containing the phrase “cardiac hypertrophy” to unearth a total of 19,718 unique biological or chemical objects that were implicitly related to the condition. After ranking those terms and eliminating a few obvious choices (like morphine), Garner and his colleagues hit upon the compound chlorpromazine (CPZ) to test in a rodent model. CPZ is used primarily as an anti-psychotic and anti-emetic, and there was “no indication in Medline that any relationship between the two has been suggested,” the authors wrote. After wet lab testing, the authors found that CPZ did indeed protect against the development of cardiac hypertrophy, a finding that is “surprising and provocative,” but nevertheless calls for further confirmation and validation.
Willis said that Etexx is currently working with UT Southwestern on eight other compounds that are currently being tested in mouse models. The key to the approach, she said, is that “we’re able to go straight from our computer analysis into an animal model.” By eliminating the need for a lab facility for early discovery, the young company saves substantial costs as well as time.
Etexx plans to develop and market its own drugs eventually, but is also looking to partner with other organizations. Pharmaceutical companies, for example, “often find a compound, and then they’re looking for the specific disease that might work,” she said. “We’re going the other way around, so as compounds come up, we would work with the innovator company for that compound and identify what type of collaboration we would want to do.”
In addition, Garner said that IRIDESCENT’s ability to perform a “cohesion analysis” makes it applicable to microarray data, so the company plans to offer microarray analysis services to help support its discovery activities. The software can take in a list of the gene responders from a microarray experiment, he said, “and present connections that are potential hypotheses — various objects that are ranked that best explain the data.”
Garner, who serves as chief scientist at Etexx, said his lab “works fairly well to capitalize not only academically, but commercially” on its work. So far, technologies from the prolific group have made their way into a number of firms, including Macrogenics, Bioautomation, and NimbleGen. Further information on some other technologies the lab is developing is available at http://innovation.swmed.edu/.