Numerate Trains Its Computer-Aided Drug Design Platform On Huntington's Disease
The path toward a breakthrough drug often starts with a new insight about the molecular cause of an illness, but only a few of these discoveries lead to new treatments. Steven Finkbeiner at the Gladstone Institutes in San Francisco has uncovered a mechanism behind Huntington’s disease, and his lab is now working with the San Bruno, CA, startup Numerate to try to beat the odds in drug development.
Finkbeiner figured out how a region of a mutant protein attacks the nerves of people with Huntington’s, a devastating neurodegenerative disorder that leads to disability and eventual death. Numerate will be looking for drug candidates to block that protein.
Their joint project is not only a test of Finkbeiner’s theories about the disease, but is also a proving ground for two different methods to improve the speed and success rate of drug discovery. The difficulty of predicting which compounds will block a disease process, without causing serious side effects, has slowed and often thwarted the introduction of new therapies.
Numerate was founded in 2007 to help solve that problem. The company uses a high-capacity computational platform to sift through the data on thousands of compounds to look for the few that might stymie a rogue agent such as the defective huntingtin protein, which acts as a nerve toxin in Huntington’s disease.
The company has been refining its “in silico” method of evaluating drug candidates as an alternative to traditional lab procedures. Under current methods, a pharmaceutical company might synthesize hundreds of small molecule drug candidates and laboriously test each in lab assays, in a “trial and error” quest to weed out the best prospects, says Numerate CEO Guido Lanza.
“We move as much work as possible to the computer, and test tens of compounds rather than hundreds or thousands, to get to the same endpoint,” Lanza says.
Numerate will be searching for a small molecule that prevents the mutant huntingtin protein from folding into a particular structure that, according to Finkbeiner’s research, helps it destroy nerve cells. Once Numerate has identified its best bets, Finkbeiner’s lab will use a second new method that may also revolutionize drug discovery. His lab will first test the drug candidates on human nerve cells grown in a lab dish, rather than in animals that have been engineered to develop a facsimile of Huntington’s disease in humans.
The nerve cells should display some or all of the abnormalities seen in Huntington’s disease, because they’ll be derived from skin cells donated by Huntington’s patients. The skin cells will first be transformed into a type of stem cell called an induced pluripotent stem cell, which can be coaxed to form nerve cells or other types of cells in the body.
Numerate will share in a $1.3 million grant recently awarded by California’s stem cell research funding agency, the California Institute for Regenerative Medicine (CIRM), to support the Huntington’s disease project.
In the search for new drug molecules, Numerate uses a statistical method similar to the way Amazon figures out what new books you’ll like based on the data it compiles about your past purchases. Numerate screens big libraries of chemical tests and other data to predict whether a family of compounds is likely to bind to a drug target such as the defective huntingtin protein, based on the compounds’ track record of affinity to other molecules. The data screening also looks at other traits of the candidate drugs, such as their tendency to form harmful interactions with other proteins.
The earliest version of Numerate’s computational drug design platform was acquired from now-defunct Pharmix of Brisbane, CA, which Lanza co-founded. Numerate’s methods were further developed during a series of partnerships with pharmaceutical companies, including Merck and Boehringer Ingelheim of Canada. Numerate has raised a total of $9.5 million in Series A and B rounds, and has received undisclosed revenues from pharmaceutical partners and government agencies, including the US Department of Defense.
The company’s scientists have created algorithms that determine which criteria should be used to weed out unpromising chemical families, and how soon those criteria should be applied, Lanza says. For example, a drug for a disease of the brain should be able to cross the blood-brain barrier. Asking that question too soon might eliminate a family of small molecules that includes at least one member of promise, Lanza says. On the other hand, asking the question too late could lead to a waste of resources on groups of chemicals that will never work for the disease in question, he says. The ultimate aim is to find a new chemical family that is not already patented for use against the target illness.
To process its mountains of data, Numerate rents computer capacity through the cloud services operated by Amazon and Google. Thus, Numerate doesn’t have to maintain its own data centers.
Among the databases of chemical information that Numerate licenses, some sets of tests date back to the 1950’s. But large amounts of information have been generated since the 1990’s, when genome sequencing and other techniques began revealing formerly unknown mechanisms of disease as fresh targets for drugs. Companies such as Exelixis of South San Francisco set up automated, high-throughput assay systems to test compounds for binding to these new targets, such as kinases—a group of enzymes implicated in cancer.
“People treated it like a land grab,” Lanza says.
Numerate’s view is that the value of this information lies in its interpretation. The data can be “noisy, biased and complex,” Lanza says. He’d like to see pharmaceutical and biotechnology companies share all that raw data, because it could help identify many new leads for the entire industry to pursue. But much of the data is still closely held in proprietary libraries, Lanza says.
In its early years, Numerate focused on providing drug design as a service. But the company is now concentrating on its own drug candidates, which offer a better promise of … Next Page »