Craif Patent Portfolio Analysis

One pulse, and a seven-year record behind it

The money arrived in a single event: $33.2M in Series D, from Granite-Integral, Kodomo no Mori and Global Brain Corporation. A round is a moment. It tells you what a set of investors concluded on one day. It tells you nothing about what the company decided to protect in 2018, when it was founded, or in the five years of filing that followed.

Craif describes itself as a biotechnology company building a cancer diagnosis algorithm from a microRNA database and machine learning. That is the sentence the round was raised against. The patent file is a separate document, written earlier, by different hands, for a different audience. It is worth reading the two side by side, because they do not say quite the same thing.

Fifteen families, five of them already solid

Fifteen families since 2018, filed into EP, JP, US and WO. Five are granted, which means five sets of claims are enforceable today rather than someday. Ten are published and pending, and the oldest of those is far enough along that a grant could land at any time. For a company seven years old, five grants is a real base, not a filing gesture.

Now read the titles. A body fluid extract containing microRNA. A device for capturing extracellular vesicles, with methods for preserving and transporting them. A biomolecule recovery device and method, plus an analysis device and method. Not one of those is a model. They are the steps that happen before the model: getting the vesicles out of urine, holding them intact, moving them somewhere, recovering what is inside.

The database is the product. The device is the property.

The filings stop where the public record does

One family in 2018. Two in 2019, three in 2020, then a quiet 2021 with a single filing. Four in 2022 and four in 2023. The curve is not explosive, but it is steady and it accelerates in the two years before the round, which is the ordinary shape of a team that has found the thing it wants to own and is filling in around it.

The last public filing is dated 20 October 2023, thirty-four months before this Series D. That gap is not an absence. Applications stay unpublished for eighteen months, so everything filed since sits in a shaded window nobody outside the company can see. Whatever the $33.2M is meant to build, the filings that correspond to it are either already lodged and invisible, or not yet written.

A small circle among very large ones

The same CPC groups, nucleic acid assays and the measuring and analysing of materials, have drawn Illumina with 504 families in five years, Zhejiang University with 490, 10x Genomics with 393, China Agricultural University with 374, Huazhong Agricultural University with 364 and South China Agricult University with 310. Against those, fifteen is a rounding error. On this chart Craif is a dot.

Which is the wrong way to read it. Count measures how much of a field an organisation touches. It says nothing about whether a specific claim is narrow and clean. In a crowded class, scope is the only thing that matters, and a device claim to a particular way of capturing extracellular vesicles from urine does not compete with five hundred sequencing families. It sits beside them.

Who has been reading the Japanese file

Citations run back to the portfolio from National Univ Tokai National Higher Education And Research System, ten of them, then West China Hospital Sichuan Univ with three, Science & Technology China University and Shanghai 3dmed Biomedical with two each, and Hope City with two. Mostly academic and hospital research, mostly in Japan and China. These are groups working the same sample-handling problem and finding Craif already standing on it.

The inventor list explains the shape. Twenty-two names across fifteen families, with Takao Yasui on eight and Yoshinobu Baba on seven. Two people carry most of the file. Akira Yokoi appears on two. That is a concentrated core doing device and materials work, surrounded by a wider bench, and it matches who is citing them: laboratories, not model builders.

What the hiring plan quietly confirms

A hundred open roles, of which five are bio or clinical and ninety-five are everything else, including device development, retail sales of a cancer risk test kit at a department store in Omiya, and internships. That is a commercialisation build, not a research expansion. The claim space is set. The company is now staffing to sell against what it already filed rather than to file something new.

So the plain version. Craif sells an algorithm and owns the sample. Five granted families and ten pending ones cover the capture, preservation, transport and recovery of the material the algorithm reads, filed from 2018 onward by a two-inventor core, in a class where the giants file hundreds of families a year and none of them need this particular device. The moat is upstream of the pitch.

The moat is upstream of the pitch.