TL;DR: On April 28, 2026, Eli Lilly and AI biotech Profluent announced a partnership worth up to $2.25 billion to develop AI-designed recombinases — custom enzymes that can perform large-scale, precise DNA edits that CRISPR cannot. Profluent's generative models are trained on the world's largest protein dataset. This is not AI augmenting drug discovery. This is AI inventing the tools of drug discovery itself.
CRISPR captured the world's imagination when it arrived. The idea that you could cut and edit DNA like a word processor — locating a precise sequence and making a targeted change — was genuinely revolutionary. Researchers won a Nobel Prize for it in 2020. Treatments for sickle cell disease followed. The biotech industry reorganized around it.
But CRISPR has a ceiling. And on April 28, 2026, Eli Lilly placed a $2.25 billion bet that a company called Profluent has found what comes next.
Why CRISPR Is Not Enough
CRISPR works by cutting DNA at a specific location. It is extraordinarily precise for small edits — correcting a single mutated base pair, disabling a malfunctioning gene, or inserting a short sequence. For many diseases, that is exactly what is needed.
But some genetic diseases require more than a cut. They require inserting a large, functional piece of DNA — sometimes thousands of base pairs long — at an exact location. CRISPR struggles here. Its editing window is small, and inserting large sequences with it is inefficient, error-prone, and in many cases simply not viable as a clinical approach.
The Problem of Large-Gene Insertion
Many genetic diseases — including certain forms of muscular dystrophy, metabolic disorders, and inherited blindness — require inserting substantial stretches of corrective DNA to restore normal function. CRISPR-based systems can typically handle insertions of a few dozen base pairs reliably. Recombinases, by contrast, can handle insertions thousands of times larger — and do so at a pre-defined genomic address with minimal off-target activity. That gap is the scientific case for the entire Profluent-Lilly deal.
Recombinases are not new. Nature invented them billions of years ago — bacteria use them to shuffle genetic material, viruses use them to integrate into host genomes. What is new is the ability to design custom recombinases from scratch, tuned to recognize a specific sequence in a specific human gene, and perform the edit reliably enough to put into a patient.
That design problem is extraordinarily hard by conventional methods. Proteins fold in three-dimensional space in ways that determine their function, and the relationship between amino acid sequence and protein behavior is not fully understood by any human expert. Exploring the design space manually — testing candidate after candidate in the lab — would take decades per target disease.
What Profluent Actually Built
Profluent is not a drug company. It is an AI company that designs proteins. Founded in Berkeley and backed by Jeff Bezos among others, Profluent built generative models trained on the world's largest database of protein sequences — including what the company describes as the most comprehensive collection of naturally occurring recombinases ever assembled.
Generative Protein Design at Scale
Profluent's models work analogously to large language models — but instead of predicting the next word in a sentence, they predict the next amino acid in a protein chain. By training on millions of natural protein sequences and their known functions, the models learn the underlying grammar of protein design. They can then generate novel sequences — proteins that do not exist in nature — with specified functional properties. For recombinases, that means: target this exact DNA sequence, insert this payload, minimize off-target activity.
The comparison to language models is more than metaphorical. In 2024, Profluent published OpenCRISPR-1 — an open-source gene editor designed entirely by AI, with no human engineer specifying the protein structure. It was the first AI-designed gene editing protein demonstrated to work in human cells. The Lilly deal is that capability applied at scale, across multiple disease programs, with a major pharmaceutical partner providing the clinical and commercial infrastructure to actually reach patients.
How the Partnership Works
Profluent designs the recombinases using its generative models. Eli Lilly receives an exclusive license to advance selected candidates through in vivo research, preclinical development, clinical trials, and commercialization. Profluent receives an upfront payment, committed research and development funding, and up to $2.25 billion in development and commercial milestone payments plus tiered royalties on net sales. Each program targets a disease with significant unmet need — the specific indications have not been publicly disclosed.
Why Eli Lilly? Why Now?
Eli Lilly is not a company that chases trends. It is one of the most consistently profitable pharmaceutical companies on earth, known for disciplined capital allocation and a long track record of successful drug development. When Lilly makes a bet of this size, it reflects genuine conviction about the underlying science.
The context matters. Lilly opened a dedicated Genetic Medicine Center in Boston in 2026 and has been systematically acquiring gene editing and gene therapy capabilities for several years. The Profluent deal fits a clear strategic thesis: the next generation of genetic medicines will require editing tools that go beyond CRISPR, and the companies that control those tools — or have exclusive access to them through partnerships — will define the field for the next decade.
The critical caveat: Milestone-heavy deals in biotech are not guarantees. The $2.25 billion figure represents a ceiling across multiple programs, not an upfront commitment. Most biotech partnerships of this structure never reach full milestone payouts — programs fail in preclinical or clinical testing, and the actual cash flow depends entirely on whether the science survives contact with human biology at scale. The deal signals conviction; it does not confirm outcomes.
AI as a Discovery Engine — Not Just a Tool
The deeper significance of this deal is what it says about the maturation of AI as a technology category.
For the past several years, most AI applications in drug discovery have been accelerants — tools that speed up processes humans already knew how to do. AI screens existing compound libraries faster. AI predicts which molecules are likely to be toxic. AI identifies patient subpopulations for clinical trials. These are valuable, but they are fundamentally AI augmenting human-designed workflows.
Profluent represents something categorically different. The recombinases it designs do not exist in nature. No human biochemist designed them. No literature describes their properties. They are outputs of a generative process that explores a design space too large for any human research program to navigate manually. The AI is not accelerating discovery — it is doing discovery.
This distinction matters enormously for how we should think about AI's long-term role in science and medicine. If AI can reliably generate novel proteins with specified functions, then the limiting factor in genetic medicine shifts from "can we design the right tool" to "can we validate and deploy it safely." That is a fundamentally different bottleneck — and one that pharmaceutical infrastructure like Lilly's is specifically built to address.
TechVernia Verdict
The Profluent-Lilly deal is a landmark moment, not just for biotech but for AI. It is the clearest signal yet that generative AI has crossed a threshold from productivity tool to discovery engine — capable of inventing solutions that did not previously exist, in domains where the stakes are as high as they get.
The path from AI-designed enzyme to approved genetic medicine is long, expensive, and uncertain. But the direction of travel is clear. The companies that learn to operate at the intersection of generative AI and molecular biology — building the models, the datasets, and the clinical trust required to deploy these tools in patients — will define the next generation of medicine. Lilly has placed its bet. Others will follow.
Frequently Asked Questions
CRISPR is a gene editing system that uses a guide RNA to locate a specific DNA sequence and a Cas protein to cut it. It is highly precise for small edits — point mutations, short insertions, gene knockouts — but struggles with large DNA insertions. Recombinases are enzymes that recognize specific DNA sequences and catalyze recombination events — cutting, rearranging, or splicing DNA — at those sites. Custom-designed recombinases can perform large-scale precise insertions that CRISPR cannot, making them essential for diseases that require inserting long stretches of corrective genetic code.
Profluent is an AI biotech company founded in Berkeley, California, focused on generative protein design. It uses large language model-style architectures trained on protein sequence data to design novel proteins with specified functions. The company is backed by notable investors including Jeff Bezos. In 2024, Profluent released OpenCRISPR-1, the first AI-designed gene editor demonstrated to function in human cells, establishing its scientific credibility before the Lilly partnership was announced.
No. In standard biotech deal structures, milestone payments are contingent on achieving specific development and regulatory milestones — such as advancing a candidate into clinical trials, completing Phase 1 or Phase 2 studies, or receiving regulatory approval. The $2.25 billion figure represents the total potential value across all programs if every milestone is achieved. Most partnerships of this type do not reach full payout because programs fail during development. Profluent does receive an upfront payment and committed R&D funding regardless of milestone outcomes.
Profluent and Lilly have not publicly disclosed the specific disease targets in their collaboration. However, the most compelling use cases for large-gene insertion technologies include: Duchenne muscular dystrophy (which requires inserting a large functional dystrophin sequence), certain inherited retinal dystrophies, lysosomal storage disorders, hemophilia A and B, and various metabolic liver diseases. These conditions share a common feature — the therapeutic correction requires delivering a substantial piece of functional genetic material, beyond what CRISPR can efficiently insert.
The Profluent-Lilly deal represents a shift from AI as an accelerant to AI as a discovery engine. Previous AI applications in pharma — virtual screening, toxicity prediction, biomarker identification — augmented existing human workflows. Profluent's generative protein design creates solutions that did not previously exist and could not have been discovered through conventional research timelines. If this approach proves reliable in clinical development, it will accelerate the design of therapeutic tools across multiple modalities — not just gene editing, but potentially antibodies, enzymes for metabolic diseases, and other protein-based medicines.
Conclusion
The announcement on April 28, 2026 was easy to read as another large biotech deal. The numbers are striking, but biotech deals of this scale happen regularly. What is not routine is what the deal represents technically.
Profluent did not discover a natural recombinase and optimize it. It trained a generative model on the largest protein dataset ever assembled and used that model to design enzymes that do not exist in nature — enzymes with specified functions, tuned to perform a specific kind of genomic surgery at a specific address in the human genome. And it demonstrated that those designs actually work in human cells.
That is a different kind of AI milestone than most people are paying attention to. The models generating headlines are the ones that write essays and produce images. But the models quietly being deployed in molecular biology labs — designing proteins, predicting structures, engineering new biological machinery — may ultimately have a larger impact on human health than anything else being built today.
The code of life is being rewritten. For the first time, AI is holding the pen. And Eli Lilly just paid $2.25 billion to be the first to use what it writes.
Related Articles:
