Materials patent enablement after Amgen v. Sanofi

After Amgen v. Sanofi, a materials patent can still claim a whole family of compositions, but only if the application teaches a skilled scientist how to make and use the entire family without trial-and-error screening. That is the core of materials patent enablement today. The breadth you can hold depends less on how many examples you file and more on whether your examples reveal a rule that predicts the rest.

This article is for materials scientists and chemistry startup founders preparing a first filing on a new polymer, catalyst, alloy or functional material. I cover what the Supreme Court held, how the Federal Circuit and the USPTO have applied it since 2023, how much data an application needs, and how I draft genus and species claims.

Key takeaways

• Breadth must be earned. Claim an entire class and you must enable the entire class. “The more one claims, the more one must enable.”
• A rule beats a pile of examples. Twenty-six working antibodies were not enough in Amgen because nothing predicted which others would work.
• Function-only claims are the weak point. Claims defined by a result (a conductivity, a yield, a strength) without structure are most exposed.
• Written description is a separate test. You must also show you possessed the claimed genus.
• Prophetic examples are allowed but must be written in the present or future tense and cannot substitute for data in an unpredictable space.
• Draft a ladder of claims from the broad, rule-based genus down to the specific composition you will sell.

What did the Supreme Court decide in Amgen v. Sanofi?

On May 18, 2023, a unanimous Supreme Court held that Amgen’s claims to a genus of antibodies defined by what they do were invalid for lack of enablement. The Court ruled that a patent claiming a class of compositions must enable a skilled person to make and use the whole class, with no more than reasonable experimentation.

Amgen had identified the amino acid sequences of 26 antibodies with the claimed functions, but the claims reached potentially millions more. Its “roadmap” for generating and screening candidates, and its method of substituting amino acids and testing the results, were in the Court’s words “little more than two research assignments.” Trial-and-error screening grants a “hunting license,” not an enabled invention.

The opinion by Justice Gorsuch in Amgen Inc. v. Sanofi is short and readable. I recommend it to any scientist who will be named as an inventor.

Why this is a materials case, not just a biotech case

The Court’s two leading precedents were materials cases.

In the Incandescent Lamp Patent case (1895), Sawyer and Man claimed lamp filaments made of any “carbonized fibrous or textile material.” Their own carbonized paper was defective. Edison later found, through painstaking experimentation, that a Japanese bamboo worked. The Court refused to let a claim to “every fibrous and textile material” cover Edison’s bamboo, noting that most fibrous materials did not work.

In Holland Furniture Co. v. Perkins Glue Co. (1928), the patent claimed any starch glue that would have substantially the same properties as animal glue. The key ingredient was described by its function, not its chemistry. The broad claim failed because others could practice or avoid it only after elaborate experimentation. The narrow claim survived.

As the Court put it, “Today’s case may involve a new technology, but the legal principle is the same.” A claim to “any polymer having a glass transition temperature above 150 °C and a dielectric constant below 2.5” sits squarely in this line.

How does materials patent enablement work after Amgen?

Enablement asks whether a skilled person, reading your application on its filing date, could make and use everything your claim covers without undue experimentation. After Amgen, the question for a materials claim is whether the specification gives a predictive rule for the whole claimed family, or only a method for finding members one at a time.

The requirement comes from 35 U.S.C. § 112(a). It requires a description that would enable any person skilled in the art to make and use the invention. Courts treat enablement and written description as separate requirements.

The Supreme Court did not demand perfection. It said that “a specification may call for a reasonable amount of experimentation to make and use a patented invention.” If the application discloses a general quality running through the class, a few examples can enable the whole class.

The Wands factors still apply

Examiners and courts measure “reasonable” experimentation using the eight In re Wands factors listed in MPEP § 2164: the breadth of the claims, the nature of the invention, the state of the prior art, the level of ordinary skill, the predictability of the art, the amount of direction provided, the existence of working examples, and the quantity of experimentation needed.

On January 10, 2024, the USPTO published its Guidelines for Assessing Enablement in View of Amgen (89 Fed. Reg. 1563). Examiners “will continue to use the Wands factors,” the guidance applies regardless of technology, and it announces no major change in practice. The notice addresses enablement only.

For materials and chemistry, the predictability factor carries a lot of weight. MPEP § 2164.03 notes that in cases involving unpredictable factors, “such as most chemical reactions,” more disclosure may be required, and that one working species may not show what other species will do.

How has the Federal Circuit applied Amgen since 2023?

The Federal Circuit has applied Amgen strictly to genus claims defined by function. Separately, it has held that the written description requirement reaches every limiting part of a claim, including a Jepson preamble. Both lines of cases matter for materials claims that combine a broad class with a performance requirement.

Baxalta v. Genentech (2023): eleven examples, millions of candidates

Four months after Amgen, the Federal Circuit decided Baxalta Inc. v. Genentech, Inc. (Sept. 20, 2023). The claims covered any antibody that binds Factor IX or IXa and increases its procoagulant activity. The patent disclosed eleven such antibodies, against millions of candidates.

The court found the facts “materially indistinguishable” from Amgen. The patent disclosed no “common structural (or other) feature” predicting which candidates would work. The court held that “random trial-and-error discovery, without more, constitutes unreasonable experimentation.” It also said it did not read Amgen to disturb the Wands factors.

In re Xencor (2025): every limiting phrase needs support

In In re Xencor, Inc. (Mar. 13, 2025), a precedential decision, the court turned to written description. It held that “the limiting preamble of a Jepson claim must be supported with sufficient written description.” A Jepson claim is a format that recites a known product in the preamble and then claims an improvement to it.

Xencor was decided on written description, not enablement, and does not turn on Amgen. But the lesson carries over: no part of a claim escapes review. If your claim recites “a battery comprising” your new electrolyte, the application must support the battery context, too.

How much data does a materials patent application need?

There is no fixed number of examples. The application needs enough data to show a pattern that a skilled person can extend across the claimed scope with confidence. In a predictable chemistry, a few examples plus a clear rule may support a broad genus. In an unpredictable one, many examples may support only what you made.

In Amgen, 26 examples did not support a functional genus. Yet MPEP § 2164.02 says a specification “need not contain an example” if the invention is otherwise disclosed well enough. Coverage and predictive value matter more than count.

Working vs. prophetic examples

A working example describes work actually performed, with actual results. A prophetic example describes an experiment that was not performed, with predicted results. Both are permitted. MPEP § 2164.02 says prophetic examples “should not be described using the past tense, but rather in future or present tense.”

I use prophetic examples to show how the rule applies to a new monomer or dopant, never to carry a claim where the team itself would not bet on the result.

Where to spend lab time before filing

Ten examples clustered around one composition say little about the edges of a range. Five placed at the corners and center say much more. I ask teams to map their data against the claim they want and fill the gaps that matter most. That planning often happens alongside the timing questions I discuss in when to file before a conference or paper.

Genus vs. species claims: how broad can materials claims be?

A species claim covers one specific material. A genus claim covers a family defined by shared features. After Amgen, a genus can be as broad as the structure-property rule your data supports, and no broader. The safest genus claims define members by structure or composition, with function as an added limitation rather than the only one.

Written description adds a second test. Under MPEP § 2163, a genus can be supported by a “representative number of species” or by common structural features that a skilled person would recognize. MPEP § 2163, citing the Federal Circuit’s en banc Ariad decision, treats written description as “separate and distinct from the enablement requirement.”

Three examples from materials and chemistry

A polymer family. Suppose you have made poly(aryl ether) films with a fluorinated side chain and measured low dielectric loss. A claim to “a polymer having a dissipation factor below 0.002 at 10 GHz” is a functional genus with no structural boundary, the Holland Furniture problem in modern form. A claim to a defined repeat unit, with the side chain selected from a listed group and the fluorine content in a stated range, ties breadth to chemistry the reader can reproduce. If you can explain why the side chain lowers loss, and your data shows the trend across several chain lengths, you have the “general quality” the Court described.

A catalyst class. A claim to “any supported metal catalyst that converts CO2 to methanol at above 80% selectivity” invites the Incandescent Lamp objection: most candidates in that class will not work, and the reader must screen to find the ones that do. A stronger claim recites a defined metal combination, support and loading range, with dependent claims for your best preparation. Selectivity then narrows a structurally defined class instead of defining it.

An alloy composition range. Composition ranges are common and usually definite; MPEP § 2173.05(c) notes that specific numerical ranges generally do not raise a definiteness issue. The enablement question is whether the whole range works. If your high-entropy alloy forms a single-phase solid solution in only part of the claimed window, the full-window claim covers brittle intermetallics too. Data near the boundaries, plus a phase-stability explanation, supports the range. Without it, claim narrower and keep fallback ranges in the specification.

A claim ladder for a new material

Each rung should stand alone if the rung above it falls. See also my pages on patent strategy for advanced materials and on chemistry and small-molecule patents.

What should a chemistry startup do before filing? A drafting checklist

Before filing, a materials startup should decide which product it must protect, map its data against the broadest claim it wants, write down the structure-property rule, and build fallback positions into the specification. Good materials patent enablement is planned in the lab, before the attorney drafts a word.

1. Name the commercial embodiment. Make sure at least one claim squarely covers the composition and process you expect to sell or license.
2. Write down the rule. State why the class works: a shared motif, a mechanism, a phase relationship or a descriptor. If the team cannot state a rule, the broad claim is probably a research assignment.
3. Map examples to the claimed space. Plot your data across the ranges you plan to claim and look for empty corners.
4. Include useful negative results. Data showing where the rule stops working justifies a boundary.
5. Define by structure first, function second. Use performance properties to narrow a structurally defined class, not to define it.
6. Label prophetic examples correctly. Use present or future tense for anything not actually performed, and keep them tied to the stated rule.
7. Disclose fallback ranges and subgroups. List narrower ranges and preferred subgroups so later amendments have written description support.
8. Describe measurement methods. Specify how each claimed property is measured (instrument, frequency, temperature, sample preparation).
9. Support the context, not just the material. If you claim the material in a device or method, describe that device or method with enough detail to support it.
10. Plan follow-on filings. If new data will arrive within months, consider a provisional application now and a fuller nonprovisional or PCT application within the 12-month window set by 35 U.S.C. § 119(e), adding data as it becomes available.

Investors ask whether the claims cover the product or only a lab curiosity. I discuss those diligence questions in what investors ask about your patents, and early-stage planning on my startups and university spin-outs page.

Can AI-generated or computational predictions support a genus claim?

Computational screening can help explain why a class works, but predictions alone are unlikely to enable a broad genus in an unpredictable field. For materials patent enablement, modeling works best when it supports a stated rule that experimental data then confirms across the claimed space, with predicted members presented as predictions.

A validated density functional theory or machine-learning model can be the “general quality” that links your examples to the genus, if the application describes the model and its validation against measured data. The weak version is a claim to every composition the model flags, backed by a handful of syntheses. That resembles the Amgen roadmap. The question is whether testing is confirmation or discovery.

Frequently asked questions

Does Amgen v. Sanofi apply to materials and chemistry patents?

Yes. The Supreme Court stated that the principle is the same across technologies, and its leading precedents were materials cases about lamp filaments and starch glue. The USPTO’s January 2024 enablement guidelines also say they apply regardless of technology. Any claim to a class of polymers, catalysts, alloys or compounds must enable the whole class.

How many examples do I need for a genus claim?

No rule sets a number. What matters is whether your examples, together with the explanation in the specification, let a skilled person predict which members of the genus will work. In Amgen, 26 examples did not support a functional genus. In a predictable chemistry with a clear structure-property rule, fewer well-placed examples may support a broader claim.

Are prophetic examples still allowed after Amgen?

Yes. MPEP § 2164.02 permits prophetic examples, which describe predicted results of experiments not performed, and says they should be written in the present or future tense. They are most useful to show how a stated rule extends to new members. In unpredictable areas, they cannot replace working data across the claimed scope.

What is the difference between enablement and written description?

Enablement asks whether a skilled person could make and use the full scope of the claim without undue experimentation. Written description asks whether the application shows that the inventor possessed the claimed invention on the filing date. Both come from 35 U.S.C. § 112(a), and the Federal Circuit treats them as separate requirements. A claim can fail either one.

Can I claim a material by its properties alone?

You can, but it is the riskiest way to claim. A claim defined only by a result, such as a conductivity or tensile strength, covers every material that achieves it, including ones you never made or predicted. After Amgen and Baxalta, I recommend defining the class by structure or composition and using properties as additional limitations.

Should I wait for more data before filing?

Not necessarily. Waiting risks losing priority to others or to your own publications. A common approach is to file a provisional application with the data you have, keep generating examples across the claimed space, and file a fuller application within 12 months. The broad claims then rest on the strongest data you can gather in that window.

Talking it through

I trained as a chemist before law, and I read a materials application the way a referee reads a paper, looking for the rule that ties the data together. If you are deciding how broadly to claim a new material, or whether your data supports the genus you want, I am happy to talk it through.

This article is general information, not legal advice, and reading it does not create an attorney-client relationship. Patent law changes and every situation differs. Please talk to a patent attorney about yours.

Message Alexander on LinkedIn