This story originally ran on June 10.
Kinexus Bioinformatics announced this week that it has developed an algorithm that accurately predicts which of nearly 500 human protein kinases are most likely to target a given phosphorylation site.
The algorithm, which was developed in collaboration with researchers in the Mathematics of Information Technology and Complex Systems groups at the University of British Columbia and Simon Fraser University, will allow Kinexus to do high-resolution mapping of an estimated 25 million kinase-substrate interactions, aiding future study of these regulatory proteins, Steven Pelech, the company's president and chief scientific officer, told ProteoMonitor.
Protein kinases are key to cellular signaling pathways and important drug targets, playing roles in over 400 different diseases including cancer, diabetes, immune disorders, and Alzheimer's disease. However, relatively little is known about the targets of the majority of these proteins. A literature search done by Kinexus researchers turned up experimentally determined substrates for just 120 of 515 human kinases and a total of 10,000 substrate-kinase interactions. Prediction matrices for likely substrates based on consensus substrate sequences exist for roughly 60 kinases and their isoforms.
"We really just don't have that much information about protein kinases, how they're regulated, what they target, how they may be connected to disease," Pelech said.
Seeking to fill some of these gaps, Kinexus researchers compiled datasets based on existing research of 10,000 kinase-substrate pairs, collecting the consensus sequences of the substrates and the primary amino acid sequence of the corresponding kinases. They applied mutual information analysis to this data to develop the new algorithm, which, Pelech said, is able to determine the kinases likeliest to phosphorylate any one of the human proteome's presently known 92,000 phosphorylation sites.
The company has run all 92,000 phosphorylation sites against the algorithm for the 500 human kinases it covers and compiled the top 50 kinases for each site, posting this information on its open access PhosphoNet database. It has also posted data from cross-species studies designed to determine which phosphorylation sites have been best conserved throughout evolution – a likely signifier, Pelech said, of which sites are most functionally important and worthy of study.
The next step, he said, is to identify the other roughly 400,000 phosphorylation sites Kinexus believes exist in the human proteome and run them against the algorithm for each of the 500 kinases, as well. This information, he said, should be available on PhosphoNet by the end of the summer.
"Our goal is to develop the first high-resolution map with 25 million kinase-substrate interactions basically quantified. And through looking at the evolutionary analysis and the scoring we can start to predict the critical connections," Pelech said. "We're able to map the architecture of what I call a molecular intelligence system for the first time, and that allows us to generate all kinds of predictive hypotheses that can be easily tested. It's like having a map of where your research wants to go."
Since launching in 1999, Kinexus has focused mainly on its in-house services business, doing work for some 1,300 laboratories around the world including 200 pharmaceutical labs. Last year the company's revenues from services were around $1.2 million, Pelech said.
Moving forward, however, Kinexus plans to expand its product development, he said. Although information generated by the new algorithm will be available for free on PhosphoNet, the company expects that access to this data will increase kinase research, and, in turn, the demand for services and reagents needed for that research.
"A lot of the research going on in the world is the same proteins always being studied while the bulk of the proteins encoded by the human genome are ignored," Pelech said. "Part of that is because the tools aren't there to study these proteins. Because of the work we've done here we can now define very nice reagents for all of the human protein kinases."
"It creates the market. The products division is now starting to open up using information from our bioinformatics and then the products are redeployed back into our services," he said. "You'll see next year that Kinexus will start to become a major products company for cell-signaling proteins with enzymes, antibodies, peptides, and microarrays all coming."
Among the specific products Kinexus plans to launch are peptide substrate microarrays and antibody arrays for kinase research – for which the company will be making many of its own antibodies. Both of these will likely become available this fall, Pelech said, along with mass spectroscopy services for the analysis of phosphorylation sites.
The company is also considering spinning off several new firms.
"We have investors that are interested in forming diagnostic spin-out companies from Kinexus," Pelech said. "We're also thinking in a few years of spinning out an early development pharmaceutical company because we have the infrastructure within Kinexus to screen, to identify drug targets, and to evaluate them. I'm in no rush to do anything, but if it seems right I will."