Ambitious Bio launches Corpus healthy-human protein atlas for AI

Ambitious Bio announced Corpus, a protein atlas — a map of which proteins are present, and in what relative amounts, across healthy human tissues — aimed at training and guiding biological AI, saying the company is backed by $6 million in seed funding.

HEALTH desk — biological AI is bottlenecked less by models than by missing maps of human proteins; a seed-stage atlas, a catalog of which proteins are in healthy tissue and in what amounts, is trying to sell that missing layer.

What the company says the map is for. The release says Corpus is meant to give drug companies and frontier labs a more complete body map for biological AI — software that tries to learn biology, not only language. Proteins, the release says, are the working parts the body actually produces in particular places and at particular levels. The genome is the parts catalog of what the body could make. It is not the list of parts one tissue is making today. Many medicines work by engaging a particular protein. That protein is the drug’s target. Healthy tissue is the baseline: scientists use it to see what disease changed, and where a medicine might act in a place it was not aimed at. File that framing as the company’s. This desk did not measure a tissue.

The inventory claim is a company comparison. In 181 head-to-head tissue comparisons against major public references and published studies, Ambitious says adding Corpus increased the detected protein inventory every time. A detected protein inventory is the list of proteins a method reports as present in that tissue. The company says the gain in a typical matched tissue ranged from 28% to 78%. Matched tissue means the same kind of tissue on both sides of the comparison. Every time, and 28% to 78%, are Ambitious’s figures. This desk did not rerun the 181 comparisons.

A second cut, still the company’s, and not the same sentence as the 28% to 78%. As one example, Ambitious compared Corpus with the Human Protein Atlas, which the release calls the field’s most widely used reference. The Human Protein Atlas, shortened to HPA, is a public map of proteins in human tissues. The company says Corpus expanded HPA’s protein parts list by 56% in the median shared tissue: 56 proteins that had not previously been identified as present in that tissue for every 100 already known. Median means the middle result, not the average and not the biggest. The company says it repeated the analysis with the next six largest references and found 79% to 138% more protein types in the median tissue that those references had not shown. File 56%, and 79% to 138%, as Ambitious’s. Do not fold them into the 28% to 78% line. They are a different cut. This desk did not open the underlying tables.

The money line stops where the page stops. The release says Ambitious is backed by $6 million in seed funding. A seed round is an early private check, before a Series A. The page does not name the investors, does not print a valuation, and does not name a customer. Do not add a Series letter past seed. Do not add a price. Do not add a buyer.

How the company says it builds the map. It says it generates purpose-built biological measurements, rather than trying to pull value out of the data exhaust of existing clinical systems. Data exhaust, here, means leftover records a hospital or a lab already produced for some other job. The release says those measurements have to be generated from physical specimens and made comparable across people, tissues, states of health, and time. They cannot simply be scraped from an existing record. File that positioning as the company’s. A method claim is not a lab tour this desk took.

The safety-use example is also the company’s. Ambitious says it looked at healthy-tissue results for 973 proteins being pursued as drug targets that do not yet have an approved medicine. For 19% of those targets, the company says HPA detected no healthy expression at all, while Corpus reported it for the first time. Expression, here, means the protein shows up in that tissue. For 77%, the company says Corpus significantly expands the list of healthy tissues where expression is detected. Significantly is Ambitious’s word. It is not a statistical test this desk ran. The point the release draws: seeing a target in more healthy tissues gives developers a wider field in which to look for unintended effects before they spend years and money on a clinical trial. A clinical trial is the test of a medicine in people. File the 973, the 19%, and the 77% as the company’s demonstration. This desk did not audit the target list.

Named voice on the release: Elizabeth Hudson, Ph.D., founder and chief executive. She says we have complete bills of material for cars, wristwatches, and toasters, down to the last screw, and we still do not have the same for the human body. She says that as AI becomes more capable of designing interventions into biology, the value of a more complete map of the system it is trying to change only increases. An intervention, here, means a deliberate change, such as a drug. File the name, the title, and those two sentences as hers, via the release. Her other sentences, and a board member’s Apple-and-China line, stay in Sources.

What is still a plan. The release says Corpus begins with healthy human protein measurements, and that Ambitious is already extending the same system from health to disease, to other kinds of molecules, and to other organisms, while increasing how fine and how deep the measurements are. That is a company plan. It is not a second dataset this desk has seen, and it is not a claim that Corpus today is a disease map.

Do not invent investors, a valuation, a customer, a drug this map approved, or an independent recount of the 181 tissues. Do not treat “world’s largest” or “world’s most complete” as a ranking this desk made. Do not treat the 28% to 78% line as the same statistic as the 56% HPA line.

Plain English for the rest of the card: protein atlas = a map of which proteins are present in a tissue, and in what relative amounts. protein = a working part a cell makes. Many drugs aim at one. relative abundance = how much of a protein is there compared with the others. genome = the catalog of what the body could make, not the parts it is making in one tissue today. target = the protein a medicine is built to engage. HPA / Human Protein Atlas = a public protein map the company used as a comparison. median = the middle result. seed = an early funding round. $6 million is the company’s figure. clinical trial = a test of a medicine in people. expression = the protein shows up. This filing is the 22 Sep launch. It is not an independent audit and not a drug approval.

PRIMARY here: Ambitious Bio’s 22 Sep 2026 PR Newswire release, “Ambitious Bio Launches Corpus — World’s Largest Healthy Human Protein Atlas to Give AI a More Complete Map of the Human Body,” stamped 09:20 ET — Tier A PRIMARY company wire. The Cambridge dateline, the map of protein presence and relative abundance, the world’s-most-complete and world’s-largest wording, the 181 comparisons, the every-time increase, the 28% to 78% typical matched-tissue gain, the HPA 56% median-tissue line, the next-six-references 79% to 138% line, the $6 million seed, the purpose-built measurements versus clinical-data exhaust, the 973-target demonstration, the 19% and 77% lines, the Hudson title and the bill-of-materials quote, and the plan to extend beyond healthy human proteins are the company’s. NOT claimed: investor names, a valuation, a customer, a Series letter past seed, that this desk reran the comparisons or counted the 973 targets, an independent world’s-largest ranking, a disease dataset already shipped, a drug approval, a stock tip, or investment advice. Distinct from the already-filed envisagenics-boehringer, abbvie-iambic-ai-drug, and fda-nams-nonclinical-rule.