Meeting subject-matter, enablement andwritten-description requirements when claims reach beyond the sequences you made
1. Introduction
Engineered proteins include therapeutic antibodies and enzymes, industrial biocatalysts, diagnostic reagents, biosensors andresearch tools. They are made by three broad families of methods: rational or structure-guided design, directed evolution (iterative mutation and selection, recognised by the 2018 Nobel Prize in Chemistry awarded in part to Frances Arnold) andcomputational or AI-driven design, which was recognised in the 2024 Nobel Prize in Chemistry awarded in part to David Baker and to Demis Hassabis and John Jumper for protein structure prediction.
From a patent perspective, these approaches share a common tension. Inventors typically make and test a small number of sequences, but want claims covering a much larger family of functional variants. U.S. law requires that the specification show the inventor possessed what is claimed and teach the skilled person how to make and use the full scope of it. This article explains how to disclose designed and evolved sequences so that claims are both valuable and defensible.
2. Patent-Eligible Subject Matter (35 U.S.C. § 101)
Natural products and laws of nature are not patent-eligible. In Association for Molecular Pathology v. Myriad Genetics (2013), the Supreme Court held that isolated naturally occurring DNA is not eligible, but that complementary DNA (cDNA), which is not naturally occurring, is. The same logic shapes protein claims.
- Wild-type proteins, even if isolated or purified, are at risk under § 101 unless they have markedly different characteristics from what exists in nature.
- Engineered variants with sequences, structures, or properties that differ from any natural protein are generally treated as eligible, particularly when the differences are non-trivial and produce new functions or properties.
- Fusion proteins, chimeras andde novo designed proteins are strong candidates because they have no natural counterpart.
- Methods of use and manufacture must integrate the natural principle into a practical application and not merely recite it.
- Computer-implemented design methods face additional scrutiny under the abstract-idea exception. Claims that tie model training and design to a specific, concrete wet-lab output, such as expressing and testing the designed protein, are in a better position than purely mathematical claims.
| Drafting tip: describe how the claimed protein differs from the closest natural sequence in structure and function andgive measured data showing the difference. This builds the “markedly different” record early. |
3. Written Description and Enablement (35 U.S.C. § 112(a))
Section 112(a) requires a written description of the invention anda disclosure that enables the skilled person to make and use it. These are separate requirements andboth are often litigated against biotechnology claims.
Written description
The specification must show that the inventor had possession of the claimed invention as of the filing date. For genus claims, courts look for a representative number of species or common structural features that distinguish members of the genus. In Ariad v. Eli Lilly (Fed. Cir. 2010, en banc) the court confirmed that written description is a separate requirement from enablement andlater decisions such as Juno Therapeutics v. Kite Pharma (Fed. Cir. 2021) held that claims defined by function alone, without adequate description of the structures that perform it, can fail.
Enablement and Amgen v. Sanofi
In Amgen Inc. v. Sanofi (2023), the Supreme Court unanimously held that claims to a genus of antibodies defined by their binding and blocking function were not enabled. Disclosing 26 antibody sequences and a roadmap for finding others, in the form of conservative substitution and a “swapping” method, was not enough to enable the full scope of a claim covering potentially millions of candidates. The Court said that the more one claims, the more one must enable.
Directed-evolution and machine-learning inventions are directly affected. A specification that says “mutate, screen andrepeat” is effectively a research plan andit risks being characterised as requiring the skilled person to repeat the inventor’s whole discovery process.
How to respond in drafting
- Disclose many working species, not only the best performer. Include variants across the spectrum of activity, so that the data supports the breadth claimed.
- Identify structure–function relationships. Map which residues, motifs, or domains are essential and which can tolerate change, supported by alanine scanning, deep mutational scanning, structural data, or model-derived attributions.
- Show the predictability of the method. If a design or evolution workflow reliably produces functional variants, present success rates, library sizes andhit rates.
- Align claim scope with data. If the evidence supports sequences within a defined identity range and containing specified residues, claim that anduse dependent claims to step back to better-supported subsets.
- Add fallback claims to specific sequences, specific mutation sets anddefined motifs, so a successful attack on the broad claim does not collapse the portfolio.
4. What to Disclose: Designed and Evolved Sequences
A specification for an engineered protein should be built around the sequence data, but sequences alone are rarely sufficient. The table sets out the core content.
| Element | What to include | Why it matters |
| Reference (parent) sequence | Source organism, database accession, full sequence andany prior-art engineered variants | Defines the baseline against which novelty and improvement are measured |
| Variant sequences | Full-length sequences with SEQ ID numbers, mutation lists in a consistent numbering scheme (such as T45A) andidentity to the parent | Supports each claimed species and the genus |
| Design or evolution method | Rational design logic, library construction (error-prone PCR, shuffling, site-saturation), selection conditions, number of rounds, and, for AI methods, model type, inputs andtraining data | Shows how the variants were obtained and supports reproducibility |
| Functional data | Activity, specificity, stability, expression levels, binding constants andassay conditions, with comparison to the parent and to benchmarks | Shows the technical effect and supports utility and non-obviousness |
| Structure and mechanism | Crystal or cryo-EM structures, homology models, active-site or interface residues andrationale for the mutations | Establishes structural features common to the genus |
| Negative and neutral results | Variants that were inactive or lost function andpositions tolerant or intolerant of change | Helps define the boundaries of the genus and supports enablement |
| Production and use | Expression systems, purification, formulation andapplications | Satisfies the “make and use” requirement |
| Biological material deposits | Cell lines or vectors deposited where sequences alone are not sufficient to reproduce the material | Provides access to material under Budapest Treaty procedures |
Sequence listings
Nucleotide and amino acid sequences disclosed in an application must be submitted in a sequence listing in the format required by the USPTO. Since July 1, 2022, applicants must use WIPO Standard ST.26, an XML-based format, in place of the earlier ST.25 text format. Sequences must be assigned sequence identifiers andthe listing is filed as a separate file that is not counted toward page limits. Practical points:
- Include every sequence that meets the length and type thresholds in the standard, including those that appear in figures, tables andexamples.
- Reference sequences in the text by their SEQ ID NO andkeep the numbering consistent between the claims, description andlisting.
- Use the WIPO Sequence tool, or compatible software, to generate and validate files before filing.
- Check how variants, unnatural residues andmodified amino acids are annotated, because ST.26 has specific rules for features and qualifiers.
- Verify the listing against the specification as part of final proofreading, since mismatches are a common source of delay and post-filing amendments.
5. Claim Strategies for Engineered Proteins
No single claim format works for every case. A balanced set covers a range from narrow and certain to broad and uncertain.
| Claim type | Example approach | Strengths and risks |
| Specific sequence | An isolated polypeptide comprising SEQ ID NO: 2 | Strong support and easy to prove; easy to design around |
| Sequence identity | A polypeptide at least 90% identical to SEQ ID NO: 2 and having a specified residue at position 45 | Covers variants; must be supported by data across the identity range; vulnerable if the range is too wide |
| Defined mutation set | A variant of a parent protein comprising substitutions at positions X, Y andZ, selected from specified residues | Concrete and often better supported than function-only claims |
| Structural motif or domain | A polypeptide comprising a defined binding loop or conserved motif | Useful for families; support depends on showing the motif’s role |
| Functional genus | Any protein that binds target X with affinity better than Y | Most exposed after Amgen and Juno; typically hard to sustain without extensive support |
| Method of making or use | Method of producing the variant, or treating a condition with it | Complementary andcan survive when composition claims are challenged |
| Related molecules | Nucleic acids encoding the variant, vectors, host cells, compositions, kits | Broadens protection across the product chain |
Cautions on identity and conservative-substitution language
- A percentage identity claim covers many sequences that were never made. Support it with data on variants distributed across the claimed range, not just clustered near the top.
- Phrases such as “conservative substitutions” should be defined in the specification, ideally with a table or a stated substitution scheme.
- Specify the alignment algorithm, parameters andthe length over which identity is calculated, so that infringement and validity analyses are not ambiguous.
- Consider product-by-process and functional limitations carefully andpair them with structural limitations wherever possible.
6. Special Issues for AI-Assisted and Computational Design
- Inventorship. Under current U.S. law, as confirmed in Thaler v. Vidal (Fed. Cir. 2022), an inventor must be a natural person. Where AI tools contribute, the human contribution to conception must be significant. The USPTO has issued andrevised, guidance on AI-assisted inventions, so confirm the current version before filing. Document who selected targets, designed constraints, evaluated outputs andvalidated results.
- Disclosure of the model. Describe the model architecture, training data, inputs, output format andvalidation in enough detail to support enablement. Where the model is proprietary, consider describing its function and results andweigh whether to deposit or describe key components.
- Experimental validation. AI-generated sequences are most defensible where the application shows they were expressed, purified andtested. Data on success rates across designs helps establish predictability.
- Non-obviousness. Examiners may argue that applying a known tool to a known protein is routine. Emphasise unexpected results, such as improvements in stability, activity, or specificity that are not predictable from the prior art.
- Trade secrets versus patents. Training sets and model weights may be better protected as trade secrets, while the sequences, methods andresults are disclosed in the application. Plan this split with care.
7. Prior-Art Searching for Protein Inventions
Sequence-based prior art is found in both patent and non-patent sources. A thorough search should include:
- Public sequence databases such as UniProt, GenBank andthe Protein Data Bank, which may disclose natural or engineered sequences that anticipate or render claims obvious.
- Patent sequence databases and commercial services that index sequences disclosed in published applications and patents, including ST.26 and legacy listings.
- Scientific literature and preprint servers, which often describe engineered variants well before patent filings.
- Structure-based and motif searches in addition to full-length sequence alignment, to find functional analogues that differ in primary sequence.
- Searches for mutation-level disclosures, since a prior reference may disclose the same set of substitutions on a homologous parent.
Run searches before drafting to shape claim scope andagain before filing and during prosecution, since new publications continue to appear.
8. Drafting and Filing Checklist
| Item | |
| ☐ | Describe the closest natural sequence and show how the engineered molecule is markedly different. |
| ☐ | Provide a representative set of full sequences with SEQ ID numbers across the claimed range. |
| ☐ | Include assay methods, conditions andcomparative data for each key variant. |
| ☐ | Identify essential residues and tolerant positions, with supporting structure–function data. |
| ☐ | Describe the design, evolution, or AI workflow in enough detail to be reproducible. |
| ☐ | Define identity, alignment method, conservative substitution andnumbering conventions. |
| ☐ | Draft claim tiers from specific sequences to broader genera, plus method and related-molecule claims. |
| ☐ | Prepare the ST.26 sequence listing and verify it against the text and claims. |
| ☐ | Consider deposits of biological material where necessary. |
| ☐ | Confirm inventorship and document human contribution, especially for AI-assisted work. |
| ☐ | Plan foreign filings, noting differences in the treatment of functional claims and sequence-based claims. |
9. Conclusion
Engineered proteins can be strongly protected, but the era of broad functional genus claims supported by a handful of examples is over. The strongest applications combine many working species, structural insight andtransparent description of the design or evolution process with claims that scale from precise sequences to carefully justified families. For AI-assisted work, the same principles apply, with added attention to inventorship and to documenting experimental validation. Planning the disclosure alongside the claims, rather than after them, is the surest way to keep that protection intact through prosecution and litigation.
This article is provided for general informational purposes and is not legal advice. Statutes, rules, fee schedules andUSPTO practice change; confirm current requirements and consult a registered patent practitioner or trademark attorney before relying on any point here.
