CellType Uses AI to Simulate Human Biology for Drug Discovery
CellType (celltype.com), a YC W26 startup, wants to replace animal models and cell lines with computational simulations of human biology. The company combines biological foundation models with AI agents that automate the drug discovery pipeline. It is backed by Pioneer Fund and other investors.
CellType was co-founded by David van Dijk and Ivan Vrkic. Van Dijk is a Yale professor with over 11,000 citations who has published in Cell, Nature, ICML, and NeurIPS. Vrkic co-developed CellType's core technology, Cell2Sentence, at Yale (published at ICML) and previously built software to control CERN's Large Hadron Collider.
The core of CellType's approach is Cell2Sentence, a method for translating single-cell gene expression data into sequences that large language models can learn from. Where models like AlphaFold focus on individual proteins, CellType aims to represent whole biological systems. The goal is something the founders describe as a "virtual human": a computational stand-in that drug developers can query at every stage of development to predict what will happen in patients, rather than relying on animal proxies.
"It's not too often that a tenured professor steps away to build a company around their own work — and work that is quite a big contribution to the field," said Dave Messina, General Partner of Pioneer Fund's Future of Health fund.
The technology has already produced a real-world result. In collaboration with Yale, Google DeepMind researchers built a 27-billion-parameter version of the model called C2S-Scale, based on Google's open-source Gemma family. Researchers screened over 4,000 compounds and identified Silmitasertib (CX-4945) as a promising candidate for enhancing immune visibility in tumors that typically evade detection. When tested in human neuroendocrine cell models the AI had never seen during training, silmitasertib combined with low-dose interferon produced roughly a 50% increase in antigen presentation. Neither drug alone achieved comparable results.
Google CEO Sundar Pichai called it "an exciting milestone" for AI in science when he shared the findings on X in October 2025, garnering over 6.8 million views. Google detailed the work in a blog post.
That result has already translated into a commercial relationship. In March 2026, Senhwa Biosciences (TPEx: 6492) signed a strategic MOU with CellType to integrate its AI platform into the clinical development of Senhwa's lead asset, Silmitasertib (CX-4945), the very compound the model had flagged during the original screen. CellType says it is also working with Top 10 pharma companies, and that all of those engagements have come through inbound interest.
The opportunity CellType is targeting is large and structurally painful. The vast majority of drugs that enter clinical trials never reach patients, and the ones that do typically take over a decade and billions of dollars to develop. Much of that failure stems from preclinical models that don't reflect real human biology. CellType's thesis is that foundation models trained directly on human cellular data can narrow that gap before expensive trials begin.