In 2006, when personal genomics company 23andMe launched from its offices in California, it charged $1,000 for a home spit kit and its top competitors were DeCodeMe and Navigenics. If you’ve never heard of those last two, that’s because they never made it out of the aughts. And for a while there, no one thought 23andMe would either.
But today, the company is doing better than ever. On Tuesday, 23andMe announced a huge new funding round that brings its reported valuation up to about $1.75 billion—big even by Valley standards. And that’s not because the company is making boatloads off of spit kits, now priced at a stocking-stuffer-friendly $99. While technically 23andMe is still in the business of consumer genetic tests, it’s time we start calling the company what it really is: a DNA data mining operation digging for the next big drug.
In 2015, 23andMe began inking lucrative research agreements with pharma giants like Genentech and Pfizer, in addition to launching its own R&D program to sift through DNA samples from its more than 2 million customers. With 85 percent of those people signing over their As, Cs, Ts, and Gs to science, it’s the world’s largest consented, re-contactable database for genetic research, according to the company. (Which is why Genentech paid $60 million for access to it.) And while 23andMe plans to put its data to work on many medical maladies, it’s always had its eyes on a cure for one in particular: Parkinson’s. This week, they showed off the latest results of that effort, turning up more than a dozen new mutations associated with the disease.
Parkinson’s is an incurable neurodegenerative disease that affects 1 million Americans. The biggest risk factor is age, but certain genetic mutations can increase risk, too. 23andMe reports two of them to its customers, which each increase the chances of developing Parkinson’s between about 30 and 75 percent. (Sergey Brin, the Google co-founder and ex-husband of 23andMe founder and CEO Anne Wojcicki, carries one of those mutations, called LRRK2.)
But scientists don’t know how the disease works, or even how much genetics play a role in its development and progression. Unlike Huntington’s or hemophilia, there’s no single genetic signal for Parkinson’s. 23andMe is betting that through brute-force pattern matching, it can map the constellation of genetic causes—and potentially inform new treatments.
So on Monday, researchers from 23andMe, Genentech, and the National Institute on Aging published the largest meta-analysis of Parkinson’s disease to date, using data from more than 425,000 individuals—more than three-quarters of whom were 23andMe customers. Building on a smaller study from 2014, the team compared genes from 6,476 self-reported Parkinson’s patients with those of more than 300,000 disease-free 23andMe customers, confirming genetic variants already associated with the disease and turning up 17 new ones.
David Hinds, a principal scientist in 23andMe’s statistical genetics group and a co-author on the paper, started at the company a little less than eight years ago. Back then, his team thought they might be able to identify most of the Parkinson’s genes with information from just a few thousand people. But they quickly realized the problem was more complicated. “These much larger sample sizes are required to make reliable discoveries,” he says—especially when you don’t understand how the disease operates. “You need a lot of findings to see how some of these associations line up along different cellular processes.”
Most of the new genes reported in the paper are involved with processes that remove old tired organelles and break down toxic proteins in brain cells. That’s not exactly news. Pharmacogenomics analyst Derek Lowe, who writes a drug development blog for Science, says the paper shows promise, but might not be more than that. “It is a pretty large dataset, but GWAS studies have their limitations in suggesting new drug targets, particularly in well-worked-on areas like Parkinson’s.” That’s because GWAS, or genome-wide association studies, are more likely to pick up common traits than mutations that directly cause disorders. Those tend to be so rare, their signal so faint, that they get missed.
Which is why Genentech’s next step is to sequence the full genomes of 3,000 of 23andMe’s Parkinson’s patients. These volunteers have answered questions about their family history, how quickly their disease is progressing, what treatments they’ve tried and how well they’ve worked. By drilling down into all 3 billion base pairs, the pharma firm hopes to get past Parkinson’s most common traits—the ones that each exert a small effect to sum up to the heritability of the disease. Instead, they’re looking for those rare variants, which wreck more biological damage machinery, leaving a trail of rubble that’s easier to track.“They’re the extreme breaks in the system,” says Rob Graham, a senior scientist in Genentech’s human genetics group, and co-author on the Nature Genetics paper.
Genentech will have exclusive access to this whole genome data for two years after the last participant is sequenced—after that, the whole scientific community can use it. But even with a head start, it’s a long road ahead. “Personalized medicine is more challenging in Parkinson’s, than in say, cancer, because of the complex heritability of the traits,” says Graham. The most promising Parkinson’s drug in clinical trials is an antibody that targets the abnormal protein bundles that build up inside nerve cells. The gene that codes for these disfigured proteins was first discovered in 1997.
A drug target does not a drug make. But at least for 23andMe, it can still make for one heck of a business plan.