“My impression is that we are currently laying the foundation for true prediction methods.”
Genedata Launches ‘Omics’ Database Module Derived from EU’s InnoMed PredTox Project
Swiss bioinformatics firm Genedata last week rolled out a new module for its Expressionist data-analysis suite that it developed through its involvement in Europe’s InnoMed PredTox consortium — a three-year, public/private effort to assess whether combining data from multiple “omics” technologies with conventional toxicology methods can help improve preclinical safety evaluation.
The new database module, called Collector, is a commercial version of a system the company developed to capture and store all the experimental data generated by the PredTox initiative, which kicked off in 2005. This database, called PredTox DB, currently houses 1.3 terabytes of gene-expression, proteomic, and metabonomic data.
Jochen Koenig, collaboration manager for Genedata, told BioInform that the company is not distributing these data. Rather, customers can use the module as a framework to capture, integrate, and analyze their own omics information.
Collector is “perfectly linked” with the data in PredTox DB, “but Genedata is not a distributor of the data content,” he said. “That data content belongs to the consortium as a whole. Eventually it will become public, and then of course we can also ship it along with our product. But until the official public release of that content, Genedata must not distribute it without the consent of the other partners.”
Like PredTox DB, Collector captures experimental data from transcriptomic, proteomic, and metabonomic platforms, and also stores related information like study protocols and gene or protein annotations. Genedata said Collector is designed to “mirror the experimental design of complex in-life studies and relationally store large amounts of experimental raw data.”
PredTox is also part of a three-year EU pilot project called InnoMed, or Innovative Medicines, that launched in 2005 with €8 million ($12.6 million) in funding from the European Commission. The consortium includes 12 pharmaceutical companies, three universities, and two technology providers (see sidebar for a complete list of participants).
Genedata is the sole bioinformatics firm in the consortium and is responsible for creating PredTox DB and for processing and analyzing all of the data the project generates.
PredTox is scheduled to run through the end of the year, and it’s unclear whether it will continue beyond next January. In April, the EC announced that it plans to fund a long-term version of the InnoMed project, called the Innovative Medicines Initiative, with a total budget of €2 billion through 2013. The EC plans to dispense around €123 million under IMI this year, but has not specified any projects that it intends to support. However, it recently released a list of 18 “scientific priorities” for the initiative.
Regardless of whether PredTox continues in its current form after January, Genedata and other consortium partners believe that it has already met its goal of assessing whether omics technologies might be able to contribute to toxicology assessment, and to determine what types of data would be the most informative for that purpose.
“The declared goal of the project is to be able, by the end of this year, to make a clear report to the pharma management: ‘This is how the different technologies performed; this is how they contributed to the more informed decision-making; this is … the effort it takes to generate the data, the feasibility of the data analysis, the throughput — all these considerations that you would have to take into account before you start building a large database,’” Koenig said.
“I’m very positive that we will be able to make these recommendations by the end of this year,” he added.
Koenig said that a “key aspect” of the project is to assess the omics technologies against each other and against traditional toxicity endpoints, “and so far we have seen very encouraging results along those lines.”
At the annual Society of Toxicology conference in March, members of the PredTox consortium presented the first results of the project, which studied 14 proprietary compounds that the pharma participants had dropped from development after they failed conventional toxicology tests. The partners also analyzed two reference compounds: the hepatotoxic diabetes drug troglitazone and the nephrotoxic antibiotic gentamicin.
The consortium researchers treated rats with the compounds at various dosages and exposure times and then analyzed kidney, liver, blood, and urine samples using a wide range of techniques: microarray-based gene expression studies; proteomics analysis with 2D-PAGE, 2D-DIGE, and SELDI; and metabonomics with liquid chromatography and mass spectrometry.
They also used conventional toxicity methods such as histopathology, hematology, clinical chemistry, and urinalysis. All of this data was compiled in PredTox DB and analyzed with Genedata’s Expressionist suite of tools.
Koenig noted that the findings presented at the SOT meeting serve as proof that omics data can help inform toxicity testing.
For example, he said, in the troglitizone study, “across the molecular levels — transcriptomics, proteomics, and metabonomics — we’ve seen a confirmation of the findings of the mechanism of toxicity, and also a very nice match with the traditional methods, so that was very reassuring.”
On the other hand, with gentamicin, “depending on the compounds you’re looking at, depending on the pathology the compound is causing, the different molecular levels can give you complementary information,” he said. Nevertheless, he noted that finding is “encouraging” because “we never had any contradicting information or situations where you would say that one technology failed completely.”
He noted, however, that it is still too early to tell whether it might be possible to rely entirely on omics data to predict whether a compound will be toxic.
“My impression is that we are currently laying the foundation for true prediction methods,” he said, but stressed that the goal of PredTox was only to prepare the groundwork for building a database that might eventually support predictive tools. Asked to predict how large a database might have to be to eventually enable predictive toxicology, “I wouldn’t want to make a guess right now,” he said.
PredTox participant Kirstin Meyer, who is a researcher in the nonclinical drug safety group at Bayer Schering, agreed that the “correlation was very good” between the omics data and traditional toxicity endpoints. Beyond that, she said, the omics data provides insights that would not be possible via traditional tox methods, such as “hints for interesting genes, or endogenous metabolites, and of course this is information you cannot get from histology.”
Ultimately, she said, the type of data captured in PredTox might help identify toxicity biomarkers that are “more sensitive than the common markers” that are currently in use, but she stressed that “this was not the primary aim” of the project.
Like Koenig, Meyer said that despite the promising results from PredTox, “it’s a little bit early” to draw any conclusions about the predictive power of omics data alone. “I could imagine that it may be possible in some cases,” she said, “but in general I would like to really have the final picture.”
She said that so far PredTox has answered a number of technical and logistical questions about the feasibility of bringing together multiple types of omics data from more than a dozen pharmaceutical firms in a single database.
“That was a very important goal, to see whether it is possible to compare data, to use certain kinds of standardization to really compare [data from] different sites,” she said. “That was something I think we have already achieved and we can show that it’s possible.”