You cannot patent a scientific discovery as such. Laws of nature, natural phenomena and abstract ideas are off limits under US patent law, no matter how new or hard-won they are. What you can patent is a specific, practical thing built with that knowledge: a machine, a process, a manufactured material, a composition. This article covers US federal law, and the facts were checked on October 10, 2026.

The line sounds tidy until you try to draw it for a diagnostic test or a gene sequence, which is where most of the litigation has happened. Here is how the courts and the Patent Office draw it today.

The short rule: knowledge is free, applications are not

The statute gives the starting point. Under 35 U.S.C. 101, anyone who invents or discovers a new and useful process, machine, manufacture or composition of matter, or a useful improvement to one, may obtain a patent, subject to the other conditions in Title 35. The word “discovers” is in the text, which surprises people. But the Supreme Court has long read the statute with three unwritten exceptions: laws of nature, natural phenomena and abstract ideas.

In Association for Molecular Pathology v. Myriad Genetics, the Court put it bluntly: a groundbreaking or even brilliant discovery does not, by itself, satisfy the eligibility test. In the 1980 decision on a human-made bacterium, Diamond v. Chakrabarty, the Court gave the classic illustration that Einstein could not have patented E=mc² and Newton could not have patented gravity. Those are, in the Court’s words, tools free to all.

The reasoning is about what a patent does. It lets the owner stop everyone else from using something for roughly twenty years. If someone could own gravity, or the correlation between a blood marker and a drug dose, every later researcher would build on top of a toll booth.

One more point trips people up. Usefulness and novelty are separate requirements. A finding can be entirely new to science and still fail at the eligibility gate, and a mundane device can pass it easily. Judges ask what the claim covers, not how impressive the work behind it was or how much money funded it.

Discovery versus invention, side by side

Usually a discovery (not patentable alone) Usually an invention (can be patentable)
A law of physics, or a mathematical relationship found in nature A device that applies the law in a new, concrete way
The sequence and location of a naturally occurring gene A synthetic molecule that does not exist in nature
A newly observed correlation between a biomarker and a health outcome A treatment method with specific steps that go beyond the correlation
A naturally occurring organism or mineral, as found A human-engineered organism or material with markedly different characteristics

The right-hand column is not a guarantee. A claim still has to be new, non-obvious and adequately described. Eligibility is only the first gate.

What Myriad did and did not decide

Myriad Genetics had isolated the BRCA1 and BRCA2 genes, which are linked to inherited breast and ovarian cancer risk, and had patents on the isolated DNA. The Supreme Court, deciding the case on June 13, 2013, held that a naturally occurring DNA segment is a product of nature and does not become eligible just because it has been isolated from the surrounding genome. Myriad had found where the genes were and what their sequence was. It had not created or altered that genetic information.

Complementary DNA, or cDNA, came out differently. Because it is built in the lab and leaves out the non-coding stretches (introns) found in the natural gene, the Court treated it as something new, and therefore eligible. The Court also noted what it was not deciding, including method claims and new applications of knowledge about the genes.

The practical lesson for a lab: the sequence is the discovery, and the engineered construct or the method of using it can be the invention.

The Mayo problem: when a diagnostic claims a natural law

Myriad is half the story. The other half is Mayo Collaborative Services v. Prometheus Laboratories, decided unanimously on March 20, 2012. The patents covered a way of tuning the dose of thiopurine drugs by measuring metabolite levels in a patient’s blood. Higher levels suggested a risk of toxicity, lower levels suggested the dose might not work.

The Court said the claims amounted to a natural law, the relationship between metabolite levels and dosage effects, plus steps that doctors and labs already performed routinely. Adding well-understood, conventional activity to a law of nature does not turn it into an invention. A claim needs what the opinions call an inventive concept, something significant beyond the law itself.

That decision is why diagnostic companies often struggle with eligibility, even when the underlying research was expensive and important. It is also why the same experimental finding can be unpatentable as a correlation and patentable as part of a new instrument or a new treatment protocol.

How the Patent Office applies the test

Examiners follow a two-part framework laid out in the Manual of Patent Examining Procedure, section 2106. First, is the claim to a process, machine, manufacture or composition? Second, is it directed to a judicial exception, and if so, does it add “significantly more”?

The second question has its own sub-steps. Examiners ask whether the claim actually recites a law of nature, natural phenomenon or abstract idea, as opposed to merely involving one. A teeter-totter relies on the law of the lever without reciting it. Then they ask whether the claim integrates the exception into a practical application. If it does, the claim is not treated as directed to the exception. If it does not, the claim moves on to a final check for something beyond routine, conventional steps.

Recent agency activity has focused mostly on software and machine learning. A precedential decision from the Appeals Review Panel, Ex parte Desjardins, found machine-learning claims eligible where they improved how a model used storage, and USPTO memos issued December 4 invite applicants to file declarations with factual evidence on eligibility, according to a law firm summary from January 2026. That summary did not state the year of the memos, so check the USPTO site for the exact date. None of this reopens the door to claiming natural laws themselves.

Congress may change the rules, but has not yet

A bipartisan bill, the Patent Eligibility Restoration Act (S. 1546 in the current Congress), would rewrite the exceptions. On July 14, 2026, the Senate Judiciary Committee held a hearing titled “From Genes to Machines: The Patent Eligibility Debate.” A Holland & Knight analysis dated July 21, 2026 reported no passage, noted limited legislative days remaining, and suggested the next Congress might be a more realistic target. It also noted that the Supreme Court has declined several recent petitions to clarify eligibility. As of today, the case law above still controls.

What this means for researchers publishing results

Timing matters as much as eligibility. Under 35 U.S.C. 102, a public description of your invention before the filing date can defeat novelty. The United States offers a one-year grace period for the inventor’s own disclosures, but many other countries do not, so a conference talk or preprint can cost foreign rights.

  1. Separate the finding (the correlation, the sequence, the mechanism) from the thing you built with it.
  2. Ask whether the application has concrete, non-routine steps, or a new physical or chemical product.
  3. Talk to your technology transfer office before publishing or presenting.
  4. Keep dated lab notebooks showing who contributed what.

If the discovery itself cannot be patented, what protects it?

Very little, and that is deliberate. Copyright covers the way you express findings in a paper, a figure or a dataset’s selection and arrangement, but not the facts underneath. Trade secret law can protect confidential know-how, such as an unpublished protocol or a culture method, for as long as it stays secret and reasonably guarded. The moment you publish, that protection is gone.

Researchers often run the two together. A lab may file for a patent on a new assay design, keep the cell-line preparation as a trade secret, and publish the underlying biology freely. Each piece sits under a different legal tool because each piece is a different kind of thing. The mechanism of a disease belongs to everyone; the kit that detects it can belong to someone.

Related questions

Can you patent a new species or a gene-edited organism?

A human-made organism can qualify, as the bacterium in Chakrabarty did. A naturally occurring organism as found cannot. Edited or engineered organisms are judged by how markedly different they are from what nature produces.

Can a mathematical formula be patented?

Not on its own. The Patent Office treats mathematical concepts as one category of abstract idea, though a claim that applies a formula in a concrete, practical way may be eligible.

Does a discovery still earn credit if it cannot be patented?

Yes, through publication, citation and prizes, and it stays free for others to build on. Patents were never the only way science is rewarded.

This is general information, not legal advice. A registered patent attorney can assess a specific invention.

Sources