Wireframe DNA sequence molecules structure mesh. Hands of robot and human touching on DNA connecting in virtual interface on future. Artificial intelligence technology concept. Vector illustration.
Credit: Iurii Motov / iStock / Getty Images Plus

Helix is providing Recursion Pharmaceuticals access to its vast clinico-genomic data to drive drug discovery innovation. Recursion will use de-identified data from Helix’s extensive network of health systems, such as comprehensive longitudinal clinical records combined with genomic data, to address multiple germline therapeutic areas. Using this data, Recursion intends to train AI models to accelerate drug discovery, develop biomarkers, and improve patient stratification strategies across a range of disease states. The partnership emphasizes combining real-world data and genomics to improve precision medicine development’s speed, efficacy, and scale.

“Multimodal data, including Recursion’s proprietary chemical and biological data combined with Helix’s diverse clinico-genomic cohorts provide a tremendous opportunity to build AI/ML-enabled drug discovery approaches,” said Matt Kinn, senior vice president of business development at Recursion Pharmaceuticals. “With these diverse datasets, we hope to drive novel therapeutic hypotheses to optimize the speed and efficacy of precision medicine development across various disease areas.”

Pharma’s most powerful supercomputer

“As the data sets generated by [Helix] from all of our health systems grew over time, we realized that [these data] could be extremely useful for certain drug discovery applications,” Hylton Kalvaria, senior vice president of life sciences at Helix told Inside Precision Medicine. “We met the Recursion team around the time of JP Morgan this year and quickly realized that there was a nice match between their goals outside of oncology and the data sets that we had that could be extremely useful for them and their AI-driven drug discovery process.”

According to Kalvaria, Helix offers an equivalent service in the non-oncology domain since Recursion signed a contract with Tempus in November 2023 for similar data exclusively for oncology.

“The life sciences business lags the health system business because you have to have a network, be generating data, and have your technology placed on all those systems to do interesting things on the life sciences side—that’s where [Helix] is now,” said Kalvaria. “While the deal with Recursion is pantherapeutic, which is why they wanted our data, the disease areas that we think have been interesting as we’ve talked to other life sciences companies are autoimmune disorders, metabolic conditions, neurology, and cardiology—those are the areas where we’ve curated specific data sets with all the variables that you would want to track conditions in those therapeutic areas,” said Kalvaria.

Kalvaria said that the collaboration between Helix and Recursion is very much related to utilizing the compute environment built with NVIDIA. Just over a month ago, Recursion announced that it had built the largest supercomputer in the pharmaceutical industry and the 35th most powerful in the world.

Ultimately, the idea is essentially to let that compute environment uncover associations between the clinico-genomic datasets from Helix and a 25-petabyte dataset that Recursion has collected proprietarily. The sheer volume of this genomic data makes even large language models appear small in comparison.

Diverse, dynamic data

The Helix program allows patients to have their whole exomes sequenced, and that data is paired with an extract of their electronic health record (EHR) data from the health system that Helix works with.

“The thing we’re most proud of here is the ability to assemble that longitudinal record for the patient,” said Kalvaria. “If you’re looking at lab values or some vital measurements over time, we have all those for that patient during their time at that particular health system. There are some other sources of these kinds of data, and, typically, they are static biobanks that you might try to tap into.”

Kalvaria said Helix’s data grows organically whenever they add a health system. So, Recursion and others who might work with Helix could be signing up to work with a dataset set to double multiple times over the next few years, which cannot be said for many datasets.

Another attractive aspect of Helix’s data is that it doesn’t come from just a single academic medical center or health system. Instead, the data comes from the ten health systems Helix is plugged into across the United States, providing a diverse mix of the types of patients and how they are seen in the clinic, which should reduce bias in data collection.

All in all, the collaboration between Helix and Recursion is an effort to accelerate drug development, which, according to Kalvaria, is a slog.

“Somebody discovers an interesting biomarker, they invent a way to test for that biomarker, it goes into trials, a drug comes out the other end, a diagnostic comes out on the market, and then all of a sudden, in a very natural way, patients are tested, which becomes real-world evidence that people can tap into—that whole process takes a heck of a long time,” said Kalvaria. “Because we have been able to set up these large-scale screening programs, we have almost solved the chicken-and-egg problem of drugs being on the market before you have interesting data sets to analyze.”

It will take some time for any drug to be realized from Recursion’s partnerships with the likes of Tempus and Helix, but it looks like they’ve gotten equipped with the best data and tools around to take a crack at it.

Also of Interest