GSK expands AI partnership with Tempus, fronting $70M for access to 'multimodal data'
GSK has been aggressively looking at AI, and in its newest venture, the Big Pharma has expanded a deal with an old partner in a bid to add to its capabilities.
The UK pharma put out word in the wee hours of Tuesday morning that it is broadening its partnership with AI outfit Tempus, which has specialized in de-identified patient data. Having worked together since 2020, GSK has paid Tempus $70 million upfront for three more years of partnership, with more investment possible and an option to extend the deal for yet another two years. This new deal, according to GSK, will focus on improving clinical trial design, speeding up enrollment and identifying drug targets.
GSK’s head of machine learning and AI Kim Branson tells Endpoints News that the original deal was centered around using Tempus’ de-identified patient data for targeted patient enrollment, citing Tempus’ large amount of data it had access to in its reported work with 40% of US oncologists in academic and hospital settings.
“It was a great way to identify patients and enroll them faster,” Branson said.
He added that at the same time, GSK has become more aggressive in expanding the company’s machine learning work as a growing operation, now at around 165 people across the US, Europe and Israel.
What makes Tempus’ dataset really unique, Branson said, is its multimodal data — or data that span different types, such as imaging or text or genetics. According to the AI chief, traditional real world evidence is usually along the lines of medications, some lab work, and diagnoses. However, that information doesn’t even begin to crack the surface of what the team is looking for — which is all the raw data. And that’s what Tempus is providing GSK.
Branson said there are no data like more data, but there are also no data like the right kind of data. As he explained:
Lots of real world evidence in cancer doesn’t give me that fine grain information that I really want for machine learning. So we need that multimodal data. I need access to the raw pixels of that pathologist slide. I don’t just need the pathologist report saying stage III carcinoma, right? I need the actual pictures because there’s so much more information that I can extract rather than the summarization.
According to Branson, Tempus’ dataset includes de-identified patient data, which provides clinical information such as the tumor sequence, pathology slides, imaging aspects, and circulating tumor DNA. On top of that, the dataset includes clinical outcomes.
“So now I’ve actually got this unique dataset where I can take all those bits and pieces and put them to machine learning models,” he added.
With GSK’s other technology deals and machine learning models, Branson said this reflects the “next expansion of GSK,” using data assets to fuel discovery efforts.
GSK’s CSO Tony Wood told media in his first call on Monday as CSO — and Hal Barron’s successor — that part of what GSK and Tempus will be doing together is initially focused on oncology, thanks to its proximity to precision medicine efforts and the size of Tempus’ dataset. That dataset is an estimated 34 times larger than the Cancer Genome Atlas and contains de-identified data for more than five million records.
And GSK is not the only Big Pharma working with Tempus. Wood noted that AstraZeneca and Eli Lilly are also collaborating with the AI biotech.
As for any plans to expand, Branson said that while the collaboration is mainly focused on cancer and identifying new targets, there is an option to go into other indications.
“Our deal is for all Tempus data, it’s not just for oncology data. So there’s some cardiovascular stuff emerging, there’s a lot of neuro emerging. And that’s also something that we’re interested in as well. And not only that, we’re interested in also working with Tempus to even expand the types of data they do collect,” Branson added.
Tempus is the biotech that Groupon co-founder Eric Lefkofsky launched in 2015 to effectively marry precision medicine and AI and bring it to physicians and researchers. And the biotech has found its fair share of top-tier backers, bringing in more than a billion dollars in its first five years. Baillie Gifford, Franklin Templeton, Google, Novo Holdings, and funds and accounts managed by T. Rowe Price invested $200 million in a Series G-2 round back in 2020.
In one of their more recent public actions together, the pair announced the start of an open-label Phase II trial back in January that would evaluate GSK’s PARP inhibitor Zejula (which was developed alongside Tesaro) in patients with advanced or metastatic solid tumors and a germline or somatic PALB2 mutation.