DCVC DTOR 2024: TechBio is emerging as the data-centric, AI-native coun­ter­weight to biotech

We use TechBio, an inversion of biotech, as a name for the new branch of the AI industry that concen­trates on creating large, robust, highly structured, proprietary biological and chemical datasets to fuel advanced computation and raise the hit rate in drug discovery. ​“We invest in AI-driven innovations that offer practical, real-world solutions, and that show the potential for substantial impact and scalability,” says DCVC managing partner Zachary Bogue. ​“We see TechBio as one of the major opportunity areas for AI applications.” 

(Of course, there is no bright line between TechBio and biotech. Neither approach is inherently superior; that’s why we range across the two. This chapter of the Deep Tech Oppor­tu­ni­ties Report focuses on companies backed by DCVC’s flagship funds. To learn about the broad scope of our biotech investing activity, visit DCVC Bio.) 

TechBio companies still do plenty of wet-lab work, but their fundamental business proposition is different from that of their biotech elders. TechBio companies aren’t built to pick apart the molecular pathways behind specific forms of neurode­gen­er­a­tion, cancer, or arthritis and engineer potential cures — a high-risk strategy that fails far more often than it succeeds. Rather, they’re built to collect the raw data that will reveal, through in silico modeling, which hypotheses and drug candidates are most worth testing. These tests then generate even more data that can be used to make even better predictions. The end result: more leads at lower cost and lower risk. 

We see DCVC portfolio company Relation Ther­a­peu­tics as a signal example of the TechBio sensibility. The company chose osteo­porosis as its first target and built an ​“osteomics” platform to generate the world’s largest atlas of genomic and tran­scrip­tomic data on bone cells from human patients. It applies machine-learning models to that data to predict which genes are most likely to contribute to the disease. It then knocks out those genes in other bone cells and measures the effects on bone miner­al­iza­tion. That assay data, finally, points the company toward possible disease mechanisms and suggests targets for intervention.

Given the right effector cells for disease modeling, the data-centric approach should work against many complex diseases, including neurode­gen­er­a­tive diseases, says Benjamin Swerner, chief operating officer of the London-based company. ​“The engine is that we’re trying to identify which bits of human biology to prosecute and why,” Swerner explains. With that why in hand, Relation can in theory avoid the classic biotech pitfall of spending tens of millions of dollars to get a single drug candidate into clinical trials, only to find out that it doesn’t work. [Editor’s note: pharma giant GSK endorsed this approach in December 2024 (after our publication of DTOR 2024) by entering a research partnership with Relation targeting fibrotic disease and osteoarthritis. The company committed up to $108 million in upfront and collab­o­ra­tion-based payments, and Relation will also be eligible for potential preclinical, development, commercial, and sales milestone payments averaging $200 million per target, along with tiered royalties on net sales of products.]

Another DCVC-backed company, Recursion Phar­ma­ceu­ti­cals, in Salt Lake City, Utah, started doing TechBio long before there was a common term for it (the company was founded in 2013, and we first invested in 2016). Recursion’s focus is on generating vast amounts of biological and chemical data — more than 50 petabytes and counting — including phenomics data showing how toxins, pathogens, genetic changes, and candidate drug molecules affect the morphology of individual cells. When the company’s statistical analyses spot novel or unexpected effects or rela­tion­ships, they can quickly design follow-up experiments to see how they relate to diseases of interest. The company calls the process a ​“virtuous cycle of atoms and bits.” ​“They have one of the biggest datasets out there, which puts them at the vanguard of the data race in biopharma,” says Bogue.

Recursion went public in 2021 and already has seven drug candidates in the preclinical or clinical testing phases, all targeting diseases with high unmet need and a lack of approved or effective therapies, such as ovarian tumors and neurofi­bro­matosis. In 2023 Nvidia invested $50 million to help Recursion speed the training of its AI models. And in May 2024, the company completed BioHive‑2, the 35^th-most powerful super­com­puter in the world (according to TOP500.org) and the most powerful cluster wholly owned and operated by a phar­ma­ceu­tical company. It comprises 504 separate Nvidia GPUs, and will be used to train Recursion’s large AI models to predict how different drug candidates will affect the human body. 

“Both Recursion and Relation are positioned to exploit the remarkable convergence we’re seeing in biology,” says DCVC general partner Jason Pontin. ​“Because of advances like DNA and RNA sequencing and machine vision there’s far more data coming from the lab. Now we have the capacity to understand that data, thanks to machine-learning models running on advanced hardware. Scientists working with machines can find medicines they would never have discovered on their own, cutting years from the drug development process.”