Beijing Farber Xintian Pharmaceutical Technology Co., Ltd.

As several well-funded companies race to make generative AI models that can design new protein therapies from scratch , one small startup has found a new use for existing AI systems: compressing large proteins to better package them in gene therapies or deliver them to hard-to-reach places like the back of the eye. The startup, called ZipBio, has raised $4 million in seed funding for the idea, it told Endpoints News exclusively. “The inspiration came from the compression of technology in the digital world,” CEO Roee Farber said in an interview. Of course, shrinking the structure of a protein and retaining its function isn’t as simple as compressing digital documents into a zip file. The startup is trying to isolate small parts of proteins that are most important for making a drug and shave off the rest. The problem is that the rest of the protein is there for a reason. Removing it can cause the entire thing to collapse. ZipBio’s solution is to use generative AI to create a new backbone that holds that key part of the protein in place. “In the past, nobody succeeded with that. Now it’s something we can do relatively trivially,” Farber said. The company hasn’t published details about how its approach works, but Farber said ZipBio is using and refining publicly available AI models for protein design, rather than creating its own. The startup was founded in 2023, began work on its lead drug just eight months ago, and is already testing it in animals, Farber said. That drug candidate builds on Regeneron’s antibody Eylea as a starting point — an eye therapy that made the company $5.89 billion last year. Eylea inhibits two forms of VEGF, a protein that can spur unwanted blood vessel growth in the eyes. But just blocking the A and B forms of the protein doesn’t fully treat everyone. ZipBio has created a compact inhibitor of VEGF-C and VEGF-D and meshed it with Eylea, using AI to create a hybrid molecule that Farber says is compact enough to travel to the back of the eye, something that a simple fusion of two full antibodies likely wouldn’t do. Clinical trials could begin as soon as 2026. The startup is also working with the French nonprofit Généthon to create a compact version of a protein that’s too big to fit into AAVs — the most common viral vectors used in gene therapy — to treat a rare disease called glycogen storage disorder III, a form of muscular dystrophy. ZipBio’s seed funding was led by NFX, a venture capital firm that invests in tech startups at a seed stage or earlier. MoreVC, which focuses on early-stage Israeli tech startups, also contributed. ZipBio’s team of 12 is currently split between Seattle and Tel Aviv, Israel.