Skip to main content
Premium Trial:

Request an Annual Quote

GSK, 23andMe Gamble on Personalized Drug Discovery With Exclusive Deal


NEW YORK (GenomeWeb) – GlaxoSmithKline is giving personalized medicine another go, with consumer genomics firm 23andMe as the centerpiece of what it calls a "genetics-driven" drug discovery plan.

As part of a new R&D strategy, GSK made a $300 million equity investment in 23andMe, a deal announced this week, to gain access to a large genotype-phenotype database and to millions of 23andMe customers who have agreed to partake in research and are willing to donate their data to science. Through the deal, 23andMe gets not only the $300 million but also the resources of a pharma giant to advance its drug development ambitions.

Making a profitable business out of developing personalized medicines isn't impossible, though the history of the field is rife with tales of pharmaceutical companies that have tried and given up.

GlaxoSmithKline has some stories to tell in this regard. In the 1990s, the company, then called GlaxoWellcome, made a significant investment into using genetics in drug development, and hired the late Allen Roses to lead that work. Under Roses' leadership, GSK researchers discovered a link between the HLA-B*15701 genotype and life-threatening hypersensitivity reactions with the HIV drug abacavir, and this became an early success story for pharmacogenetics. But a decade later, the company got rid of its genetics group and Roses. 

The company at one point tried to compete with Genentech's Herceptin, the poster-child for personalized medicine, with its own HER2-targeted drug Tykerb (lapatinib), but failed. Then, four years ago, GSK decided to sell its oncology drugs to Novartis, which industry observers saw as a move away from targeted medicines, which are most amenable to personalization based on genetics.

Now, GSK is once again embracing genetics in drug development, though there are risks in the latest deal for both parties. For GSK, the pivot toward what it calls a "genetics-driven" R&D strategy may yield promising targets in the four to five years it will be working with 23andMe. However, without a long-term, institution-wide commitment to creating an ecosystem that supports personalized medicine, which relies heavily on molecular testing for diagnosis, disease monitoring, and for predicting whether patients will respond or experience adverse events, GSK may not be able to reap the value of the drugs it develops.

"They built up [their personalized medicine] capabilities, and then they blew it up," said Peter Keeling, CEO of test commercialization agency Diaceutics, who at one time worked at GSK. "There has always been this R&D-led resistance to genetic targeting of drugs."

Meanwhile, for 23andMe, a new player to the drug development space, it is significant that the terms of the deal restrict it from working with any other group on new drug target discovery projects. According to Keeling, this is a risky move, because the use of big data in drug development is still in its infancy, and no one company has all the capabilities to make sense of it.

A long-term vision

Industry insiders who have followed GSK closely are pleased to see the company re-embrace genetics in drug development, in line with other players in the pharmaceutical industry, but they also said it remains to be seen if the company has a long-term vision around personalized medicine. For what it's worth, Hal Barron, who became GSK's chief scientific officer at the start of the year and who has been charged with shaking up R&D operations, indicated that the partnership with 23andMe, particularly its database of genetic/phenotypic relationships, could be "transformational" for drug development and the key to identifying high-value targets.

In outlining its R&D restructuring plans this week, GSK said it was shifting to a "genetics-driven" versus and "genetics-supported" R&D portfolio, since drugs that are targeting a mechanism of action with a strong genetic rationale have a twofold higher probability of success compared to drugs that don't. With 23andMe's help, GSK is also hoping to pinpoint subpopulations of patients that are more likely to respond to treatments, and to more efficiently recruit patients to clinical trials.

During a conference call to announce its partnership with 23andMe, Barron acknowledged that while GSK has a good track record of launching new therapeutics — with four new products approved in the last 10 months and another approval expected soon — the company has lagged in terms of sales. "What GSK has done very well is gotten a large number of new medicines approved," Barron said. "The critique that I think is fair is that the value of each of them has been less than the industry average."

According to GSK, it has more than 40 new molecular entities in its pipeline, several of which are slated to yield pivotal data between this year and 2020. In that time frame, the company is expecting to launch two new HIV dual treatments and an antibody-drug conjugate for multiple myeloma.

In the discovery phase, Barron's plan is to focus on areas of biology that will provide opportunities to pursue multiple indications for each molecule, and the data from 23andMe will be a key part of that strategy by enabling phenome-wide association studies. "With human genetics, not only can you find a target for a disease … but you can actually ask a different question, which is, if you have a given gene, what are the different phenotypes you might see from having that polymorphic variant?" Barron said.

The first project GSK will work on with 23andMe will be focused on an LRRK2 inhibitor it has in preclinical development for Parkinson's disease. Mutations in LRRK2 are known to cause Parkinson's, and the protein resulting from the mutation "is known to be an overactive tyrosine kinase," Barron said. "By inhibiting that, we think we might be able to benefit those patients with Parkinson's that have this mutation."

Although the LRRK2 G2019S mutation is a genetic determinant of Parkinson's most commonly observed among those with Ashkenazi Jewish or North African ancestry, in the Parkinson's disease population as a whole, the mutation occurs in 1 percent of sporadic and in 4 percent of familial cases. "Even though Parkinson's is the second most common [type] of neurodegenerative disease … we would have to genetically screen 100 patients with Parkinson's to identify one or two people [with the LRRK2 mutation]," Barron said.

23andMe's customers from its consumer genomics business — around 4 million of whom have consented to partake in research — could make it easier for GSK to identify these subgroups of rare patients. For example, 23andMe has more than 250 patients with Parkinson's who also have the LRRK2 mutation, and 3,000 people with the mutation who don't have disease symptoms. "What would have taken many, many years to conduct a trial [for this indication], could be done significantly faster," Barron said.

GSK also plans to recruit clinical trial participants from 23andMe's research-consented customer base. Individuals who agree to participate in research through 23andMe often answer survey questions about their lifestyle, the drugs they're on, other conditions they have, and even the zip code in which they reside. GSK can use this information to reach out to individuals who fit certain clinical trial criteria and are located near study sites. 

"Let's say based on the zip code, we know a large number of people will be available for a clinical trial in a given area, it actually helps us decide what sites to identify to enable those people to be even closer to a site," Barron said. "Rather than just pick an institution and hope there are patients nearby, we can know where the patients are and pick an institution that's convenient for the patient. So, in some respects, its putting the patient at the center of the clinical trials process."

In addition to mining 23andMe's database, GSK said, it would also invest in sequencing capabilities, in machine learning to interpret genetic data, and in techniques, such as CRISPR, to understand the function of genetic markers. 

23andMe has amassed 1.5 billion data points on health and traits from its customers. "We hope to be able to use novel and innovative machine learning, neural maps, and artificial intelligence-types of data analysis to supplement the more routine genome-wide association and phenotype-wide association studies we do now," said Richard Scheller, who joined 23andMe in 2015 to start its therapeutics group and who worked with Barron previously at Genentech.

GSK's investment in 23andMe follows in the footsteps of Roche's play for oncology software firm Flatiron Health, noted Keeling. A few years ago, Roche led a $175 million funding round for Flatiron and said that the two companies would work together to "accelerate clinical trials, advance personalized medicine, and enhance patient care." This year, Roche announced it would purchase Flatiron for $1.9 billion. 

The 23andMe collaboration suggests to Keeling that GSK is pivoting to follow the rest of the pharma industry in embracing a R&D strategy that focuses on molecular drug targets. "That's good," Keeling said. "But the not so good news is, I don't see a commercial predisposition toward the testing market."

Each year, Diaceutics issues a Personalized Medicine Readiness Report, in which it tracks the investments and changes that top drugmakers have made to insure commercialization of their personalized medicine assets. A key measure in the report looks at whether pharmas are enabling access to the diagnostics that are critical to patients getting the right drugs.

In the most recent report, GSK ranked pretty low in terms of its personalized medicine readiness. "This is bizarre," Keeling said, since GSK markets lots of drugs for hepatitis, HIV, tuberculosis, for example, that rely on diagnostics. While GSK has made diagnostics investments periodically in some of these areas, he noted that the company has yet to articulate a long-term strategic vision for the role that diagnostics will play in its new "genetically-driven" drug development plans.

Keeling recalled that Diaceutics worked with GSK on its testing program for abacavir, during which the drugmaker showed great commitment to supporting a diagnostic strategy. "They drove HLA-B*15701 testing from an obscure test … to a test that was offered to 40,000 treatment-naïve HIV patients," he said. "They did an extraordinary job. They did publications and health economic studies and demonstrated their capability to think of the diagnostic as a key strategic component."

If GSK is hoping that its R&D restructuring will bring billion-dollar drugs into its portfolio, then it can't afford to ignore the diagnostics portion of the market. Data collected by Diaceutics has shown that incorrect, inconclusive, and delayed test results for 13 cancer-linked biomarkers are hindering around 78,000 patients per year from receiving personalized, targeted treatments. Additionally, because of inadequate investments to ensure faster turnaround times, lower false-positive rates, and sufficient sample quality around a dozen biomarker tests, pharma is losing $8.3 billion in annual drug revenue in the US, and double that amount when five leading European markets are factored in. 

As pharma R&D teams at GSK are preparing these targeted therapies for the market, Keeling advised they "start to think much earlier about the clinical testing ecosystem that will ensure the right patients are tested at the right time."

A risky bet

Since launching its therapeutic division in 2015, 23andMe has been pretty quiet about its drug development work. Following the announcement this week, 23andMe remained tight-lipped about the targets it will pursue within the collaboration.

During a conference call on Wednesday, Scheller noted broadly that the company has discovery programs for autoimmune disorders, cardiovascular disease, osteoarthritis, liver disease, and cancer immunotherapy. But he also said that 23andMe hasn't shared the details of its discovery projects with GSK yet and would disclose it to the drugmaker this week.

Without knowing these details, GSK has placed enormous faith in the data amassed by 23andMe and has even given the company certain advantages in the terms of their deal. For example, 23andMe can share in the profits from drugs developed in collaboration with GSK but also explore targets on its own that it isn't advancing in the partnership. But perhaps most important is the fact that in joining up with GSK, 23andMe gains access to a large pharma players' target validation, manufacturing, and commercial operations.  

The terms of the deal with GSK also contain some limits for 23andMe's fledgling therapeutics group, particularly in that it restricts 23andMe from pursuing new projects around drug target discovery with another group for the duration of the partnership. According to 23andMe, this stipulation doesn't impact its existing collaborations with drugmakers that involve target discovery, and the company said it will continue to provide data and analyses to academics and researchers in other areas.

However, Keeling sees the exclusivity terms as a high-risk move for 23andMe. "The use of big data and artificial intelligence … we're just at the beginning of all of that, and it's messy," he said. "We're dealing with incomplete datasets at the moment."

23andMe may have millions of research-consented customers and 1.5 billion data points, but "nobody has the entire picture," Keeling reflected. "To bet drug development on incomplete data sets with one company, that's a tall order."