On January 20, 2026, Isomorphic Labs and Johnson & Johnson’s Janssen Biotech unit announced a cross-modality, multi-target research collaboration — the first time Isomorphic has extended its AI drug-design engine beyond the small-molecule work that defined its earlier deals with Novartis and Eli Lilly. The financial terms were not disclosed, but the scope is notably broader: instead of chasing a single validated target, the two companies are pointing Isomorphic’s models at a slate of what the companies describe as historically hard-to-drug biology across small molecules, biologics, peptides, and molecular glues.
A Different Kind of AI Deal
Most AI-pharma tie-ups to date have focused narrowly on one modality — usually small-molecule chemistry, because that is where generative models have the most training data and the clearest track record. The Janssen collaboration is structured differently: Isomorphic will handle in silico compound design and prediction using its AI-first platform, while Johnson & Johnson leads the experimental validation and downstream development. That division of labor lets Isomorphic apply the same predictive engine — built on AlphaFold 3’s structural-biology foundations and what the company calls planet-scale compute — across biologics and peptides as well as conventional chemistry, categories where AI-native biotechs have made far less public progress than in small molecules.
Why “Hard-to-Drug” Matters
The industry has spent decades accumulating a graveyard of validated disease targets that nobody has been able to drug with a conventional molecule — proteins with flat, featureless surfaces, or biology that only responds to large, structurally complex therapeutics. Molecular glues and targeted protein degraders are one of the few chemistry classes to have cracked open some of these targets in the last decade, and they are notoriously difficult to design rationally because they require predicting three-way interactions between a drug, a target protein, and a cellular disposal machinery. That is precisely the kind of structural prediction problem Isomorphic’s AlphaFold lineage is built for, which is likely why J&J chose this collaboration to test the platform against biologics rather than restrict it to easier small-molecule wins.
Isomorphic’s Growing Partnership Stack
The Janssen deal is Isomorphic’s third major pharma partnership, following collaborations with Novartis and Eli Lilly that, combined with this one, are reported to be worth as much as $3 billion. Each deal effectively acts as an independent, real-world test of whether Isomorphic’s platform generalizes beyond the narrow chemistry classes it was first validated against — and each partner brings different disease areas and modalities to stress-test the system.
Reasons for Caution
Because financial terms were not disclosed, outside observers cannot judge how much conviction J&J actually has in the collaboration versus how cheap an option it was to acquire. Biologics and peptide design are also areas where public benchmarks for AI performance are much thinner than in small-molecule chemistry, meaning there is less independent evidence that Isomorphic’s models will perform as well outside their original AlphaFold-adjacent comfort zone. Some biotech analysts have noted that “cross-modality” partnerships are also a hedge for pharma companies: rather than committing capital to a single therapeutic bet, J&J is buying broad optionality across many possible molecules, most of which will still fail.
What to Watch
The real signal will come from which specific targets and modalities J&J chooses to advance out of this collaboration and how quickly. If Isomorphic’s platform can nominate viable biologics or molecular-glue candidates against previously undruggable targets within a similar 12-to-18-month window to its small-molecule work, it would be the strongest evidence yet that generative AI drug design is not confined to one chemistry class. If progress stalls, it will bolster the argument that AI’s edge in drug discovery so far is narrower than the hype around deals like this one suggests.
The Bigger Picture for J&J
For Johnson & Johnson, the Isomorphic partnership is one piece of a broader push to bring computational drug design into its Janssen R&D organization at a moment when the company, like most large pharmaceutical players, is under pressure to refill its pipeline as patents on existing blockbuster products approach expiration. Rather than building comparable AI capability entirely in-house, J&J has opted to buy access to a platform it judges to be ahead of what its internal teams could replicate quickly — a common pattern across the industry in 2026, where even pharma companies with substantial data-science budgets are choosing to partner with specialized AI-native firms rather than compete with them directly. That calculus could shift if internal AI capabilities at large pharma companies mature faster than expected, but for now it explains why deals like this one keep proliferating across the sector.
