IBM and Baylor College of Medicine have been working together to develop software that scans the scientific literature to make predictions, Technology Review writes.
The researchers put the tool to work on the more than 60,000 scientific papers mentioning the cancer-related protein p53. From its scan, the algorithm identified proteins that it deemed likely to be kinases that regulate p53.
"We have tested 10," Baylor's Olivier Lichtarge said at a cognitive computing meeting at IBM's Almaden research lab, according to Tech Review. "Seven seem to be true kinases."
Lichtarge adds that although the algorithm was written to search for kinases, it also appears able to tease out phosphatases.
He and his colleagues say that a tool like theirs could help speed up the time it takes to develop drugs and it can quickly look through new additions to the literature and make connections.
Lawrence Hunter, the director of the University of Colorado Denver's Center for Computational Pharmacology, tells Tech Review that confirmation studies are needed, though he notes that literature-mining tools are also needed.