ToposBio – unlocking “Undruggable” Targets – welcome to boldstart
Unlocking “Undruggable” Targets – boldstart’s Investment in Topos Bio
At boldstart, we partner with technical founders tackling hard problems for the autonomous enterprise. The autonomous enterprise now extends into the physical world, as AI has made previously unsolvable problems tractable, from robotics to manufacturing to techbio.
When we met Ryan Zarcone and Amir Khosrowshahi, it was clear they were our kind of people. Both are keenly focused on being at the cutting edge of their discipline and making that work useful to people.
Ryan previously led AI at Fountain Therapeutics, and Amir had co-founded Nervana, the first AI system on a chip company, acquired by Intel where he became CTO of AI for eight years. Ryan and Amir met through their work at the Redwood Center for Theoretical Neuroscience at UC Berkeley.
With their shared expertise across therapeutics and complex AI systems, in 2024 they became convinced that technology had progressed to the point where it was possible to model disordered proteins, changing drug development and disease treatment. Our initial conversation centered around the impact neurodegenerative diseases had on the lives of us and our loved ones.
Today, we’re excited to announce our inception investment in Topos Bio.
Protein Modeling: The Most Important Problem Everyone Thinks is Already Solved
We love teams who are re-inventing categories people take for granted. Most people think proteins are a solved problem, thanks to AlphaFold and others, but with today’s technology, drugs have only been designed to target ~500 proteins of the 20,000 in the human body and the ~3,500 directly linked to diseases.
That’s because current modeling techniques only work well for static structures. Roughly one-third of the human proteome is intrinsically disordered and can’t be targeted today. These dynamic proteins are behind some of the biggest diseases of our time: Alzheimer’s, Parkinson’s, ALS, and aggressive cancers like prostate cancer.
To model these proteins, you need to define an ensemble of conformations vs. one “best case” option. What if we could do that to accurately model IDPs, and then design drugs to target them?
Introducing Topos-1
Topos-1 is the first step on that journey. The model generates tens of thousands of all-atom conformations in minutes, roughly 1,000x faster than conventional molecular dynamics simulations. Topos-1 outperforms AlphaFold, Chai-1, Boltz, BioEmu, and specialized IDP models across every experimental benchmark. Topos-1 is built on a proprietary physics-based simulation engine as well as internal experimentation data obtained through SAXS and circular dichroism. You can read more in the technical report.
We’re thrilled to co-lead Topos Bio’s inception round alongside Threshold Ventures, with participation from Neo and exceptional angel investors. This is exactly the kind of deep technical innovation we believe will redefine what’s possible. Read more at Axios and on the Topos Website.
