Anto Biosciences Builds the First Foundation Model for the Gut Microbiome

Drug development has long been hobbled by a quietly devastating problem: the same medicine that works for one patient can prove ineffective – or even toxic – for the next. Increasingly, the suspect is the trillions of microbes living inside us. Anto Biosciences, a Y Combinator Fall 2025 company, believes microbiome drug development can finally be made computable.

Founded in 2025 by Arvid E. Gollwitzer and David de Gruijl, Anto is building a foundation model for microbial communities – the kind of large, generalizable AI system that has reshaped language and vision, now turned on the human gut. The company’s tagline is simple but ambitious: A Foundation Model for Microbial Communities. Its goal is to predict drug toxicity and efficacy across diverse populations, and to “fix” drugs so they work for everyone, addressing what the founders describe as the hidden root cause of most drug failures.

The pair are second-time founders with deep credentials. Gollwitzer comes from the Broad Institute of MIT and Harvard, where he pioneered quality-aware, goal-directed sparsification techniques and published in Nature Portfolio. De Gruijl trained in gastroenterology at Harvard Medical School and previously worked at Janssen Pharmaceuticals (J&J) and ETH Zurich. Together, they’ve published their breakthrough work at leading machine-learning conferences as well as in Nature Portfolio, signaling that Anto’s approach is grounded in peer-reviewed science rather than hype.

The market opportunity is enormous. Roughly nine of every ten drug candidates that enter clinical trials never reach patients, with billions of dollars lost to late-stage failures. Many of those failures trace back to interpatient variability that current preclinical models cannot capture. By treating the microbiome as a learnable system – one whose interactions with drugs can be predicted in silico – Anto aims to help pharmaceutical partners triage candidates earlier, design dosing regimens that account for microbial diversity, and ultimately deliver therapies that work across populations rather than just the narrow slices represented in trials.

Anto’s positioning sits at a fertile intersection of categories the company tags itself with: Generative AI, Biotech, Genomics, Drug discovery, and AI. That overlap matters. The current generation of AI-for-drug-discovery startups has focused mainly on protein structure or small-molecule design; comparatively little compute has been pointed at the microbial ecosystem that mediates how those molecules behave inside a real human body. Anto is betting that this gap is where the next leap in personalized medicine will come from.

Although the company is early – a four-person team operating out of San Francisco – its scientific lineage and YC backing position it to attract both the talent and the partnerships required to validate the model at scale.

If Anto succeeds, the implications are profound: a future in which physicians don’t just prescribe a drug, but prescribe a drug calibrated to the microbiome of the person about to take it. In an industry where small efficacy gains can save millions of lives, building the first multimodal foundation model for microbial communities – one that combines genomics with metabolomics to understand not just what is there, but what it does – may turn out to be one of the most consequential AI bets of the decade.