Katharina Volz is on a mission to solve one of medicine’s most intractable problems: a cure for Parkinson’s disease. But, unlike the scores of researchers before her who have tried – and failed – to find a treatment for the complex neurodegenerative disorder, Volz is taking a new approach: she’s making a map.
“Part of the problem is that Parkinson’s is really poorly understood, and it’s a really complex disease so it’s difficult to get a full picture of what’s actually happening,” says Volz, founder and CEO of medical startup OccamzRazor. She hopes that by using artificial intelligence to map everything we know about Parkinson’s, her startup will be able to fill the gaps in our knowledge and home in on potential cures.
In 2015, Volz had just finished her PhD in stem cell biology at Stanford when she received a phone call with the news that a close relative had been diagnosed with the disease. “It kind of shook my world upside down,” she says. “I was completely devastated, I stayed at home for two days and cried.” And then she got to work. “I made this resolution that I was going to find a cure for Parkinson’s disease.”
But before Volz got started on finding a cure, she had to work out exactly what we already know about the disease. She wanted to build a system that could bring together tens of millions of datasets – scientific papers, clinical trial data, patents, patient records and genomics data – and automatically mine that information to reveal new insights about Parkinson’s.
“We knew we had to build a natural language processing system just to get the information out of the format that it’s in,” Volz says. To do that, she turned to Stanford’s AI laboratory, which helped the startup develop an artificially intelligent system that could take sentences from medical literature, correctly label different genes, proteins and metabolites, and identify the relationship between those labelled substances.
OccamzRazor’s system then combines all of this data into a 3D graph that Volz calls the Parkinsome – a complete map of everything we currently know about Parkinson’s. As well as mapping the genetic components of Parkinson’s, the Parkinsome also details all of the known proteins, cell types and metabolites that play a role in the disease, and the drugs known to interact with them.
The first challenge the Parkinsome is being applied to is finding out whether there are any drugs being used for other conditions that could be repurposed for Parkinson’s. Once possible treatments are identified, Volz plans to partner with pharmaceutical firms to take the candidate drugs through the costly clinical trial process and to market. “I don’t want to wait another ten, 20 or 30 years to make this happen,” Volz says. “I want to bring effective medications to Parkinson’s patients as soon as possible.”
But it might not be a case of throwing artificial intelligence at the data. Often, the scientific literature around a disease is contradictory, or sometimes outright incorrect. Volz says her system can identify contradictory results and remove them from the Parkinsome, but there are also things about Parkinson’s that we still don’t know, and no amount of searching will turn up information that isn’t there to begin with. “There are clearly gaps in knowledge, for sure,” Volz admits. “Sometimes we [will] need to run an experiment to fill the gap.”
The company – which counts Google AI’s lead Jeff Dean as an investor, and is advised by Nobel Prize winning cell biologist Randy Schekman – won’t share its platform directly with researchers, but it is working with the Michael J Fox Foundation to identify new areas that could lead to fruitful research.
For Volz, the Parkinsome is just the beginning. The OccamzRazor database already includes information on a vast range of conditions which she says could benefit from a similar AI-led approach. “We are looking at one disease at a time, talking all of the knowledge there is and connecting it with each other.”
More great stories from WIRED
🤕 No-deal Brexit would trigger a huge data problem
🔓 The iOS 13 privacy settings you should change right now
🔋 Tesla has created a battery that could last one million miles
🗓️ Four-day working weeks aren’t the utopia they seem