10x Science raises $4.8 million to build AI for protein characterization

The real bet behind AI for proteins is not prettier charts or bigger models. It is whether software can take a dense mass spectrometry readout, identify molecular forms and chemical modifications, and hand a scientist an answer in minutes instead of forcing weeks or months of manual interpretation. That is the problem 10x Science is trying to solve after raising a $4.8 million seed round led by Initialized Capital.

The round was announced April 22 and was oversubscribed. Y Combinator, Civilization Ventures, Founder Factor and a group of strategic angel investors also backed the company, which says its platform delivers automated, explainable molecular insights in minutes while current tools and manual workflows often take months. That matters because 10x Science is not chasing a generic life-sciences assistant. It is aiming straight at a brutal bottleneck in biologics development, where specialized scientists still spend serious time combing through results from highly accurate but notoriously complex mass spectrometry experiments.

10x Science was founded in December 2025 by David Stephen Roberts, Andrew Reiter and Vishnu Tejus, who worked together in Carolyn Bertozzi’s lab at Stanford University. The founders came out of an environment where the frustration was practical and immediate: existing tools did not let researchers understand what was happening at the molecular level in immune-cell and cancer-cell interactions. That background helps explain why the company’s starting point is not a broad omics dashboard but a system built around the realities of spectrometry data, traceability and regulatory use.

The platform combines deterministic chemistry and biology algorithms with AI agents, and 10x Science says it had to do substantial work to train models on spectrometry data and keep the analyses traceable for compliance. The company says its architecture reasons across hundreds of thousands of spectra, which is the kind of scale that can turn protein characterization from a specialist slog into a repeatable workflow. That is the lane where the money is likely to follow first: drug development, especially biologic drugs, where every candidate has to be characterized at the molecular level to prove it is safe, effective and manufacturable.

The broader significance is that investors are backing a very narrow scientific wedge rather than a general-purpose protein model. AI’s most famous protein success story has been structure prediction, but 10x Science is pitching one layer downstream, where measured molecules have to be interpreted, validated and defended. With tens of thousands of biologic drugs in active development worldwide and regulatory pressure on characterization rising, that may be the part of protein AI that gets paid for first.