Diagnostics · Surgical margins
Clear Gene
An RNA test built to tell a surgeon whether the whole tumor came out. Eight years after Y Combinator, the science lives in a granted US patent and one federal grant, with no registered trial, no peer reviewed paper and no FDA record behind it.
The weak link
A lumpectomy ends with a question nobody in the room can answer. The surgeon takes out the tumor with a rim of healthy tissue around it, the specimen goes to pathology, and days later a report says whether cancer cells were sitting on the inked surface. If they were, the patient goes back under anesthesia.
What makes that report shaky is how little of the specimen anybody looks at. A lab routinely examines three to five microscopic sections cut from tissue whose entire outer surface is the thing in question. Clear Gene’s patent puts it in one sentence: “Histopathology has been the best way to examine tumors for over a century, but it is not an ideal way to hunt for residual disease on the surface of a specimen.”
It cites a review of 1,201 lumpectomy margins at Brigham and Women’s Hospital, published by Tang and colleagues in 2015, in which microscopy came in at 51% sensitivity and 69.5% specificity. Of 462 positive results, only 149 had residual tumor. “The 313 false positives triggered an alarming number of unnecessary surgeries,” the patent says.
How often margins come back positive depends on the number you take. A National Cancer Database analysis of 707,798 lumpectomy patients found 5.02%, while noting that “Previous studies suggest the rate of positive surgical margin (PSM) after lumpectomy for breast cancer is approximately 20 %.” The gap between those figures is, in a sense, the market.
Three genes
Clear Gene’s answer is to stop looking and start amplifying. Press the cut surface of the excised tissue against a coated glass slide, a technique pathologists call a touch prep, then run RT-qPCR on whatever comes off.
The panel is small: three messenger RNAs, MMP11, IBSP and COL10A1, chosen because they are expressed far more strongly in tumor than in the breast tissue beside it. The patent reports cross validation on 939 RNA sequencing samples at 96% accuracy, 96% sensitivity and 94% specificity, and an independent set of 75 at 97% accuracy. Those numbers come from The Cancer Genome Atlas, a public sequencing archive, analysed on a computer. They are not results from patients whose margins were tested with a kit.
A founder who argues about uncertainty
Brandon Steelman registered Clear Gene, Inc. in California on July 10, 2014, incorporated in Delaware, listing himself as chief executive, secretary and registered agent, with John Kresge as chief financial officer. NIH records him as principal investigator, title FOUNDER/CEO. The 2025 patent names two inventors, Steelman and Julia Meyer.
In November 2016 he published a correspondence in Nature under the affiliation “Clear Gene, Moss Beach, California, USA,” arguing that the US election polls had been fine and the models built on them had not. “We cannot claim to make highly certain predictions from highly uncertain data,” he wrote. It reads like a manifesto for a margin test, and it is the only item in PubMed carrying a Clear Gene affiliation. The company reached Y Combinator in Summer 2018 and is still listed as Active there, though it appears in neither of TechCrunch’s roundups of that Demo Day.
The pandemic detour
By October 2020 the cancer work had moved down the page. The careers section read: “Our team was selected by the NIH RADx initiative [1] to scale national COVID testing.” It set a deadline too: “Our mandate is to increase the number of SARS CoV-2 tests from ~0.5M per day to +5M per day by mid-September, before the influenza pandemic hits in the fall.”
What shipped was MICA, on a separate site: an isothermal SARS-CoV-2 kit, a lysis kit meant to replace column and bead extraction, and cloud software that reads amplification curves off other manufacturers’ instruments. The site says “Clinical validation was performed by an independent clinical lab.” That lab is not named. MICA never reached the US market as an authorized test: the FDA list of molecular emergency use authorizations for SARS-CoV-2, checked on September 23, 2026, mentions neither Clear Gene nor MICA. The product pages carry a European authorized representative and the line “For In Vitro Diagnostic Use. Available in selected countries.”
Still there, and quiet
The trail does not end, it thins. The MICA site was last republished on May 23, 2025, the patent granted on April 15, 2025, and the main site was republished on April 8, 2026, still leading with residual disease. The jobs page carries a scam warning and the line “We are not currently hiring for any roles.” Tracxn counted four employees in August 2026.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| NCI SBIR Phase I | R43CA233107, Molecular Analysis of Residual Breast Cancer, $293,237, June 2019 to January 2021. PI Brandon N Steelman. USAspending puts it at $293,236.71. The only federal award to the company. | Public record |
| SBIR Phase II | None. No R44 or follow on award in NIH RePORTER or USAspending as of September 23, 2026. | Not found |
| Granted US patent | US 12,275,994 B2, filed June 22, 2018, granted April 15, 2025. Assignee Clear Gene, Inc. Claims the MMP11, IBSP and COL10A1 panel. | Public record |
| Patent family | AU 2018428853 B2 granted and active. EP 3810807 and CA 3103572 pending. PCT WO2019245587A1 ceased. Two earlier applications drew notices of allowance in 2021. | Public record |
| Test performance | 96% accuracy on 939 cross validated samples, 97% on an independent 75. Derived from archived sequencing data, not from tested patient margins. | Company-stated |
| Clinical trials | No registered study naming Clear Gene or the margin test. | Not found |
| Peer reviewed papers | One PubMed record carries a Clear Gene affiliation, the 2016 Nature correspondence on polling. Nothing on the margin test. | Not found |
| FDA | No 510(k), no PMA, no device registration or listing, no enforcement record in openFDA, and no emergency use authorization for MICA. | Not found |
| RADx selection | The site said in 2020 that the team had been selected by NIH RADx for a $30M+ partnership. No RADx award to Clear Gene appears in RePORTER or USAspending. Some RADx work ran through subcontracts that never surface as direct awards, so absence is not proof, but the money trail stops at the 2019 grant. | Differs from the record |
| MICA COVID line | Sold with a European authorized representative and the note “Available in selected countries.” Validation attributed to an unnamed independent lab. The software is claimed to “Reduce errors up to 97% compared to IVD instrument software” against one assay on one instrument. | Company-stated |
| Own CLIA laboratory | The company sells to clinical labs, and benchmarked MICA against samples “previously tested by a US CLIA-certified laboratory.” No CLIA certificate held by Clear Gene was found. | Not found |
| Corporate standing | California document 3692774, filed July 10, 2014, formed in Delaware, active and in good standing at the July 15, 2025 registry extract. | Public record |
Read plainly: the idea is sound, the intellectual property is real and freshly granted, and the clinical evidence does not exist yet. A three gene panel that separates tumor from healthy breast in archived sequencing data is a hypothesis, not a test. The step that would turn it into one, running the panel across real lumpectomy specimens and comparing the result with what pathology found, has not been registered, published or filed with a regulator anywhere the public can see it.
What to watch
- Any prospective margin study. A single site series against histopathology would move this company further than another patent.
- EP 3810807 and CA 3103572, both still pending, and whether the lapsed PCT matters to the family.
- Whether MICA is retired. Its site has not been republished since May 2025, while the parent site was refreshed in April 2026.
- Hiring. Four people and a closed jobs page is a holding pattern, not a launch.
In their words
“We cannot claim to make highly certain predictions from highly uncertain data.”
Brandon Steelman, Clear Gene, Nature correspondence, 2016 · Peer reviewed
“Meanwhile, detection of residual disease for solid tumors continues to rely on microscopic sectioning, which is not an ideal way to hunt for residual disease.”
Project narrative, NCI SBIR award R43CA233107, 2019 · Public record
“Some pathologists therefore conclude that margin analysis is the weak link in breast cancer care.”
US patent 12,275,994 B2, Clear Gene, Inc., 2025 · Company filing
“But patients with solid tumors like breast cancer have been left behind. Until now.”
Clear Gene website, accessed September 2026 · Company
“This is an ambitious $30M+ public-private partnership with few historical precedents.”
Clear Gene careers page, October 2020 archive · Company
“We’re a group of careful scientists, quietly working behind the scenes to arm those labs on the front lines of patient care.”
Clear Gene website, accessed September 2026 · Company
Related companies
Sources
- Public recordNIH RePORTER 1R43CA233107-01A1, abstract and narrative
- Public recordUSAspending.gov award R43CA233107
- Public recordUS patent 12,275,994 B2, Methods and compositions for the analysis of cancer biomarkers
- Public recordEP 3810807 A1 file and family legal status
- Public recordIn Vitro Diagnostics EUAs, molecular tests for SARS-CoV-2
- Public recordCalifornia Secretary of State business registry, Clear Gene, Inc., document 3692774
- Peer reviewedTrump: polls right, models wrong
- Peer reviewedRates of margin positive resection with breast conservation, NCDB
- IndependentClear Gene company profile, deal history and patent list
- IndependentThe two Y Combinator S18 Demo Day roundups
- CompanyClear Gene website, home and jobs pages
- CompanyClear Gene careers page with the RADx statement
- CompanyMICA test kit, MICA Direct and MICA Software pages
- CompanyClear Gene directory entry, Summer 2018 batch
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
