Stanford University

The collaboration is in line with Incyte CEO Hervé Hoppenot’s comment to Fierce Pharma in January that the company will be considering deals that cover “new technologies that we don’t have—so, early-stage, exotic stuff.”\n Incyte is sticking to its strategy this year of finding “early-stage, exotic stuff”—paying Genesis Therapeutics $30 million upfront to use the biotech’s artificial intelligence platform to develop small molecule medicines against undisclosed targets.Genesis’ platform, nicknamed GEMS, is designed to produce novel molecules and predict properties such as potency, selectivity and pharmacokinetics. The Burlingame, California-based company has previously linked up with Genentech, Gilead Sciences and Eli Lilly, with the latter deal carrying a potential total value of $670 million for up to five targets.“AI has the potential to redefine how we discover small molecule medicines, and our team is at the forefront of this revolution,” Genesis CEO Evan Feinberg, Ph.D., said in a Feb. 20 release. “We are pleased to establish this world-class partnership to combine our GEMS AI platform with Incyte’s deep expertise and track record in drug discovery and development, with the shared goal of advancing critical treatments for patients with severe diseases.”Now, Incyte wants in, offering $30 million upfront for a collaboration focused initially on two targets in severe disease. The Jakafi maker will have the option to expand to an additional target for an undisclosed additional fee. Genesis could be in line for up to $295 million per target in development, regulatory and commercial milestones, alongside tiered royalties. “Partnering with Genesis Therapeutics presents a unique opportunity to leverage their AI technologies to accelerate the discovery of breakthrough small molecules for high-impact targets in our pipeline,” Incyte’s head of R&D Pablo Cagnoni, M.D., said in the release.The collaboration is in line with Incyte CEO Hervé Hoppenot’s comment to Fierce Pharma in January that the company will be considering deals this year that cover “new technologies that we don’t have—so, early-stage, exotic stuff.”Hoppenot also said in the interview that he was looking forward to an eventful 2025, after Incyte recovered from some bad news in its pipeline. The most disappointing development the company faced was perhaps a preclinical toxicology discovery from INCB000262, the crown jewel of Incyte’s $750 million buyout of Escient Pharmaceuticals. As a result of the safety signal, clinical development of the MRGPRX2 inhibitor was put on hold in November, less than six months after the acquisition.Genesis was born out of the Stanford lab led by Professor Vijay Pande, Ph.D., which focused on applying machine-learning-powered simulations in chemistry, biology and medicine. The biotech raked in a $52 million series A in 2020, followed by a substantially bigger $200 million series B haul in 2023.

The Trump administration’s respective nominees for CMS administrator and NIH director, Mehmet Oz and Jay Bhattacharya, disclosed this week that they will divest dozens of investments in biopharma and medtech companies if confirmed by the Senate for their new roles. Oz, a television personality, said in his disclosure that he will resign as an advisor to iHerb, a supplement company, which he previously valued between $5 million to $25 million, and as advisor to Housey Pharma, a botanicals company, from which he said he valued the role at $500,000 to $1 million. He added he would resign from his position with SandboxAQ, an AI company, which he valued at between $100,000 to $200,000, and as advisor to Eko Health, a company developing a digital stethoscope. Further, Oz said he would divest his $50,000 to $100,000 in UnitedHealth Group, $50,000 to $100,000 invested in AbbVie, and $1,000 to $15,000 in Eli Lilly, among others. The announcement of the divestments come as newly-confirmed HHS Secretary Robert F. Kennedy, Jr. has vowed to root out corruption in government and the revolving door between industry and HHS. Kennedy previously said he would divest his stakes in the biotech companies CRISPR Therapeutics and Dragonfly Therapeutics. Bhattacharya pledges to resign from Stanford University, where he was an investigator in its Center for Demography and Economics of Health and Aging. His divestments are significantly less than Oz. Mainly, he vows to resign from his positions at Collateral Global, a UK-based charity focused on non-pharma-related interventions used in the pandemic response, and the Sphere Institute, a California-based nonprofit policy research firm. He said he will also leave Biosafety Now, a non-governmental organization that advocates for reducing the numbers of high-level biocontainment laboratories, among others. Bhattacharya also wrote that he “ceased creating content for compensation” on the social media platform X and “ceased sharing in revenue for content creation for X under @DrJBhattacharya handle.”

Nearly a year ago, bioengineering professor Patrick Hsu asked: “Is DNA all you need?” Hsu, co-founder of the Arc Institute, posed that question while helping create Evo, an AI model trained on 2.7 million genomes of prokaryotes and phages. Chasing the grand hope of building AI models that start to understand and manipulate biology, Evo opened the possibility that training a model on tons of DNA — and just DNA — may be a promising path forward. After all, Evo generated a novel CRISPR system from scratch based on the genomes it was trained upon. Hsu’s team is back to take that year-old question to the next level. Its newest model, called Evo 2, was trained on far more As, Gs, Cs, and Ts: roughly nine trillion base pairs of DNA from over 128,000 different species of bacteria, archaea, and eukaryotes. On a call with reporters, team members described Evo 2 as the largest publicly available AI biology model to date. The team, which includes leaders from the Arc Institute — a nonprofit founded in 2021 to bankroll scientists — and the chipmaking giant Nvidia, open-sourced the model, and on Wednesday posted a 65-page preprint describing how the team built Evo 2 and some of the results. Evo 2 is a sweeping research collaboration with roughly 50 authors listed on the paper, most of whom are affiliated with the Arc Institute, Nvidia, UC Berkeley and UCSF. The research consortium shows how many leaders building AI biology models come from outside the biopharma industry. The author list includes engineers affiliated with AI tech startups like Liquid AI and Goodfire. Greg Brockman, the co-founder and president of OpenAI, spent some of a sabbatical working on part of Evo 2’s design that allows it to consider an extra-long length of DNA code. The term “foundation model” has become a ubiquitous buzzword in the AI world, but Evo 2 fits the narrower definition of being a model trained on tons of data that is general enough to handle many tasks. The new preprint described some of those uses, like being able to predict if certain mutations to the BRCA1 gene could cause disease. The Evo 2 team reported the model appeared to understand some biology concepts on its own, like transcription factor binding sites, protein structure elements and intron-exon boundaries. Like a true foundation model, it remains to be seen where Evo 2 could have the most impact. That contrasts with other top AI models that were designed to master one job, like AlphaFold2 being built to predict the structure of proteins. Instead, Evo 2’s builders envision researchers building task-specific applications on top of the model, and Evo 2 acting like an operating system. “From predicting how single DNA mutations affect a protein’s function to designing genetic elements that behave differently in different cell types, as we continue to refine the model and researchers begin using it in creative ways, we expect to see beneficiary uses for Evo 2 we haven’t even imagined yet,” Arc Institute’s chief technology officer Dave Burke said in a statement. Stanford assistant professor Brian Hie said he believes Evo 2, paired with other models, could open new ways of manipulating genomes. “We kind of want to go beyond designing things at the molecular level because that’s all that we could control,” Hie said on the media call. “If we have a powerful model that lets us generate things at the scale of complete organisms, then this unlocks a lot of downstream tasks for which I don’t think a lot of people have even imagined what the potential use cases are.”