For patent attorneys, few technology areas demand closer attention right now than artificial intelligence (“GenAI”). Two reports from the World Intellectual Property Organization (“WIPO”) bracket a two-year window: the published in 2024 (the “2024 Report”) and the published in 2026 (the “2026 Report”). Together, they offer a data-driven view of how AI patenting has evolved and what that evolution means for innovators and the attorneys who advise them.

The 2026 Report highlights the accelerating pace of global AI patenting, a pace that commands the attention of attorneys advising technology-adjacent clients. WIPO’s 2026 Report updates the 2024 Report, documenting a field that has entered what it calls “a new phase of technological maturity and commercial urgency.”[1] GenAI inventions roughly doubled between 2024 and 2025, rising from approximately 18,862 published patent families to 37,808 total[2], more than 56,000 new GenAI patent families were published in 2024 and 2025 alone,[3] exceeding the entire cumulative output of the preceding decade.[4] GenAI now accounts for 8.7% of all AI patent activity, up from 6.1% in 2023. These are not speculative filings. While patent filings do not equate to enforceable IP rights, and many of these applications will face validity challenges before they mature into granted patents, the sheer volume reflects genuine trends in inventiveness. This is a technology area where companies are actively innovating, building IP portfolios, and signaling their intent to compete at the frontier.[5]

Geography Matters[6]

The geographic story is the one that matters most for strategic planning. China remains firmly at the forefront, with over 43,000 patent families published in 2024 and 2025 alone, more than its entire cumulative output over the preceding decade. But the growth is becoming more global. Japan has risen dramatically to third place among inventor locations, overtaking the Republic of Korea, driven largely by SoftBank’s extraordinary filing activity. The United States posted a 92% compound annual growth rate between 2023 and 2025, with patent families rising from about 1,185 in 2023 to over 4,380 in 2025. Among companies, the most striking shift is that SoftBank has emerged as the single largest GenAI patent holder with nearly 3,000 families, virtually all filed in 2023 and published in 2025. The traditional Chinese leaders, Tencent, Ping An, and Baidu, remain among the largest cumulative holders, while Alphabet has become the leading U.S.-based patent owner with 1,083 families. The sheer concentration of activity in East Asia signals that companies operating primarily in Western markets should be asking hard questions about their competitive positioning.

It is worth noting what the 2026 Report’s data reveals about the nature of these filings. International patent families, meaning patents filed in two or more jurisdictions, remain a small fraction of total GenAI activity, accounting for approximately 9% of all filings in 2025. The Report attributes this partly to the geographic composition of the landscape: Chinese applicants typically file most of their patents domestically, and the relatively young age of many GenAI portfolios means applicants have not yet extended protection internationally. This raises a forward-looking question that is underappreciated: as Chinese and Japanese companies mature their portfolios and begin nationalizing applications in commercially significant jurisdictions, companies that have not built competitive IP in overlapping technology spaces may face meaningful exposure through licensing demands, design-arounds, or outright freedom-to-operate challenges.

The Importance of Data Strategy[7]

According to the 2026 Report, “the most widely discussed GenAI trend of 2025 has been the rise of agentic artificial intelligence (agentic AI), or systems that can autonomously plan, execute multi-step tasks and adapt their behavior based on real-time feedback loops, with minimal human oversight.”[8] What sets agentic AI apart from earlier GenAI tools? Rather than simply responding to a single prompt, these systems break down complex goals into step-by-step actions, tap into external tools and data, and refine their own outputs along the way. The field is still young, but early patent filings are already emerging, with Alphabet (Google) and Nvidia among the first movers. As agentic AI matures, expect it to claim a growing share of the GenAI patent landscape.[9]

One practical consequence of this multimodal evolution is that the value of diverse training data becomes more critical, not less, as systems grow more capable. The 2026 Report documents how the emergence of highly efficient models from companies like DeepSeek has demonstrated that competitive reasoning performance can be achieved at a fraction of the costs previously assumed, while the broad release of open-weight models has democratized access to frontier-level GenAI. But having access to model architecture in the abstract is not the same thing as holding defensible rights to high-quality, domain-specific datasets suited for a particular commercial application, especially in life sciences and health care. Indeed, we may still be in something of a data gold rush: the computational infrastructure exists, the model architectures are increasingly commoditized, and the differentiator for any given AI application increasingly comes down to whether the developer holds access to proprietary, curated, and legally defensible training data. For patent counsel, this means that IP strategy cannot stop at the model layer. Clients need to think about how they protect and control the data pipelines that feed these systems. It is worth noting, however, that the legal tools for protecting training data remain imperfect. Trade secret protection, while often the most natural fit for proprietary datasets, presents novel enforcement challenges and requires planning to identify and protect the secrets well before they are misappropriated. Companies should consider approaching data protection as a layered strategy combining trade secret protocols, contractual restrictions, and technical access controls, recognizing that no single legal theory provides complete coverage.

The Life Sciences Story[10]

With respect to life sciences and health care, one of the most telling data points in the 2026 Report is how little space the sector occupies in an otherwise comprehensive analysis. The molecules, genes, and proteins data mode, which most directly captures life sciences innovation, showed uneven recent growth: 523 patent families in 2023, a dip to 397 in 2024, and a recovery to 766 in 2025. The Report notes that while this remains a relatively small category, it “continues to reflect the growing application of GenAI in drug discovery, protein design and the life sciences more broadly[11].” Looking forward, the Report observes that “the integration of GenAI into sectors such as healthcare, manufacturing, finance and energy is likely to expand the diversity of actors filing patents in this field.”[12] Patent activity in this space has not collapsed, but it has not kept pace with the explosive growth seen in text-based and image-based AI applications.

Life sciences AI patenting remains an area worth watching rather than a dominant theme in the current landscape. The connection to multimodal systems and data quality is direct and consequential for this sector: in healthcare and drug discovery, the stakes of flawed or biased training data are exceptionally high. As multimodal models increasingly integrate clinical records, imaging, genomic data, and literature within single architectures, the provenance and quality of each data source compounds in importance. For life science applications, where a biased algorithm could misdirect a drug candidate or misclassify a patient population, the quality and provenance of training data is critical.[13] While these risks are logical and well-grounded in the technical realities of multimodal AI, it should be noted that courts and regulators have not yet addressed data quality failures in AI-driven life sciences through enforcement actions or published liability findings. The legal theories under which such failures might give rise to actionable claims remain undeveloped, making this an area where the practical risk outpaces the legal framework.

Strategic Implications

For patent attorneys advising clients in this environment, strategic imperatives are straightforward but urgent. The 2026 Report states plainly that “the patent landscape is evolving faster than ever, and the strategic decisions being made now, about where to file, what to protect and how to navigate an increasingly crowded field, will shape competitive positions for years to come.”[14] Counsel should be benchmarking client portfolios against the competitive landscape described above, paying particular attention to whether claim language drafted around older GAN-specific architectures still captures the way innovation is actually happening now that LLMs and multimodal systems have become dominant.

Of particular consideration for U.S. innovators is the fact that patenting AI technologies is significantly more challenging in the United States than in many other jurisdictions due to the restrictions imposed by 35 U.S.C. § 101.[15] Under the Alice/Mayo framework, AI-related inventions face heightened scrutiny for patent eligibility, with claims frequently characterized as directed to abstract ideas implemented on generic computer hardware. The Federal Circuit has not drawn a clear line for AI method claims, and patent examiners apply the two-step eligibility test inconsistently. This creates an asymmetry: while Chinese, Japanese, and Korean innovators can secure patent protection for AI inventions under comparatively more permissive eligibility standards, U.S. applicants face a materially higher bar. Indeed, USPTO Director John Squires has acknowledged that “categorically excluding AI innovations from patent protection in the United States jeopardizes America’s leadership in this critical emerging technology.”[16] The result is a potential competitive disadvantage for U.S. innovators who limit their patent strategy to domestic filings. If the only patents in your portfolio are those that survived § 101 scrutiny in the United States, you may be leaving significant protection on the table in jurisdictions where those same inventions would be patentable without difficulty.

As a consequence, international filing strategy deserves fresh and urgent attention. While international patent filings remain only 9% of total GenAI filings, many portfolios have not yet been extended internationally. Filing in key jurisdictions before prior art density increases represents a genuine window. For U.S. companies in particular, global patent planning is not merely optional, it is a strategic necessity. Innovators who do not plan globally risk finding themselves unable to enforce against competitors in the very markets where AI patents are most readily obtainable, while simultaneously facing freedom-to-operate challenges from foreign portfolios that mature and enter Western jurisdictions. And for any client whose AI application depends on proprietary datasets, counsel should be advising on the full spectrum of protections available for that data, from trade secret protocols to contractual restrictions on downstream use, because in a landscape where model architectures are converging, proprietary data represents a significant competitive advantage.

---

[1] 2026 Report at Introduction

[2] .2026 Report at page 11.

[3] 2026 Report at page 20.

[4]  2026 Report at page 7.

[5] Id.

[6] 2026 Report at pages 13-16.

[7] 2026 Report at page 10.

[8] Id.

[9] Id.

[10] 2026 Report at 19.

[11] Id.

[12] 2026 Report at 20.

[13] See Abulibdeh et al., , PMC (2025) (finding that 95.5% of 691 FDA-cleared AI/ML devices lacked demographic transparency and only 1.6% had randomized trial evidence).

[14] 2026 Report at 8. See also, Congressional Research Service, , IF12563 (updated Jan. 16, 2026) (noting that “some stakeholders worry that AI inventions may be at risk under the current framework because they may be characterized as methods of organizing human activity, mental processes, or mathematical concepts”).

[15] See HGF, (2025) (comparing the EPO’s “technical character / technical effect” approach with the U.S. abstract-idea exclusion at the eligibility stage). 

[16] See

Introduction

In our companion article, Pharmaceutical Method-of-Use Claims After Teva v. Eli Lilly and In re Xencor: Written Description, Enablement, and the Known-Compound Genus, we analyzed two recent Federal Circuit decisions that offer complementary guidance on a recurring question in patent law: what must a patent specification disclose to satisfy written description and enablement under 35 U.S.C. § 112 when the claimed invention is the new use of a genus of known compounds or compositions?

In Teva Pharm. Int’l GmbH v. Eli Lilly & Co., 172 F.4th 1367 (Fed. Cir. 2026), the Federal Circuit reversed JMOL of invalidity for claims directed to reducing the incidence of, or treating, headache with humanized anti-CGRP antagonist antibodies. The court held that where the composition genus is well-known in the art, the specification and evidentiary record establish that a person of ordinary skill in the art (POSA) would know how to make or obtain genus components, and the record supports a finding that all claimed members work for the specific therapeutic use, written description and enablement may be satisfied even without exhaustive disclosure of every genus member. The court credited jury findings that were supported by Lilly’s own Inter Partes Review (IPR) admissions, expert testimony, and documentary evidence, all of which established that the antibody genus and humanization methods were well-established.

In In re Xencor, Inc., 130 F.4th 1350 (Fed. Cir. 2025), the Federal Circuit affirmed a written-description rejection for claims directed to “treating a patient” with modified anti-C5 antibodies bearing Fc substitutions M428L/N434S. The court held that broad, unlimited treatment language that is untethered to any particular disease, patient population, or outcome expands the written-description burden beyond what a sparse specification can bear. The applicant’s failure to establish through credible evidence that the anti-C5 antibody genus was well-known, combined with the absence of any treatment data or working example, proved fatal.

Together, these decisions establish that the validity of method-of-use claims employing known compounds turns on specificity, evidentiary support, and standard of review. This article distills the practical drafting and prosecution lessons that flow from that framework.

Drafting and Prosecution Lessons

1. Claim the New Use with Specificity

Practitioners should identify the disease or condition, patient population, dosing or route where relevant, and therapeutic outcome or biomarker. They should avoid undifferentiated “treating a patient” language unless the specification truly supports all patients and all diseases. The contrast between “reducing incidence of or treating headache” and “treating a patient” was not cosmetic. It determined whether the court evaluated the disclosure against a specific therapeutic use or against treatment across all patients and diseases.

2. Document Known-in-the-Art Status

If the compound or composition genus is known and not itself the invention, practitioners should document that status in the specification and prosecution record with concrete citations and examples. The specification should include representative species, commercially available materials, known sequences or deposits, and methods of making or modifying the compositions. Teva’s specification cited prior-art antibodies and methods, and Lilly’s own IPR statements corroborated their well-known status.

3. Explain Why the Known Genus Works in the New Method

The specification should include mechanism, common functional property, data, prophetic examples, or rationale demonstrating why all claimed members work for the specific use. Teva turned on the finding that all humanized anti-CGRP antagonist antibodies treat headache. Xencor failed because the specification did not establish that any anti-C5 antibody, whether modified or otherwise, actually treated any patient.

4. Avoid Relying Solely on Claim Format

Jepson format makes the preamble limiting and requires written description support for the alleged prior-art context. Under Xencor, a patentee cannot obtain a Jepson claim with a preamble asserting prior art if the specification and record do not establish that the prior art is, in fact, well-known. Means-plus-function format under § 112(f) narrows a limitation to disclosed corresponding structure and equivalents, as provided by MPEP § 2181, but practitioners must actually disclose that structure and should not assume that § 112(f) solves written description problems elsewhere in the claim.

5. Preserve the Evidentiary Record

Admissions that a genus is well-known may help in validity disputes but may also be used by adversaries. Lilly’s IPR statements that anti-CGRP antagonist antibodies were well-known undermined its written description challenge. Practitioners should build a record on POSA knowledge, routine methods, and whether all claimed members work for the claimed use. They should also be aware that broad statements in IPR or prosecution may be deployed later in litigation or vice versa.

Repurposed Drugs

For repurposed or repositioned drugs, these cases offer cautious support for method-of-use claims where the drug or compound class is known and the new use is narrow and well-supported. The Teva framework applies most naturally to the drug-repurposing context, where the molecule is known and the invention is the discovery that it treats a different condition.

Several key considerations improve the possibility of success. The specific therapeutic indication should be identified with particularity. The specification should provide mechanistic rationale for why the known compound works in the new indication. The application should include in vitro, in vivo, clinical, or mechanistic evidence demonstrating the new therapeutic use. The compound’s known-in-the-art status should be documented with citations to prior art, commercial availability, or known sequences and structures. Finally, practitioners should focus on claiming what the evidence supports, because “treating a patient” without disease limitation invites the same problem that proved fatal in Xencor.

Repurposed-drug patents do not excuse failure to describe the claimed therapeutic use. Broad treatment claims require broad treatment evidence.

AI-Assisted Method-of-Use Inventions

As drug discovery continues to evolve with the application of artificial intelligence and machine-learning models, such models will increasingly be used to identify new therapeutic uses for known compounds. Where AI predicts that a known compound class is effective for a specific disease, the resulting patent application faces the same written description and enablement requirements that Teva and Xencor illuminate.

Based on the teaching of Teva and Xencor, patent drafters should address the following considerations. The AI methodology should be disclosed in sufficient detail to make predictions credible, including model type, training data characterization, and validation approach. The biological rationale supporting the AI prediction should be explained, tying computational output to mechanistic understanding. The application should provide representative experimental confirmation or robust prophetic examples that demonstrate the predicted therapeutic effect. Claims should track no broader than the prediction plus evidence reasonably supports so that the application establishes possession and enablement of the claimed method. Finally, practitioners should document that the compound genus is known, if it is, with the same rigor that Teva’s specification and record demonstrated for anti-CGRP antagonist antibodies.

Applications relying on AI predictions alone, without experimental confirmation or strong mechanistic reasoning, may face meaningful risk under the Xencor rationale. The ARP found that listing disease classes without data was inadequate, and AI-predicted indications without confirmation data would likely face the same conclusion.

Conclusion

Teva v. Eli Lilly and In re Xencor together establish that method-of-use claims employing known compounds occupy a spectrum. At one end, claims directed to a specific, well-supported therapeutic use of a demonstrably known composition genus can survive written description and enablement challenges even without exhaustive disclosure of every genus member. At the other end, claims directed to undifferentiated “treatment” of any patient with any disease, using a genus whose well-known status is merely asserted, will fail.

The practical lesson is clear: specificity in claiming and evidentiary rigor in establishing POSA knowledge are the twin pillars of valid method-of-use claims for known compounds. Practitioners who heed both decisions will be better positioned to draft claims that survive validity challenges in prosecution and litigation alike.

Introduction

Two recent Federal Circuit decisions offer complementary guidance on a recurring question in patent law: what must a patent specification disclose to satisfy written description and enablement under 35 U.S.C. § 112 when the claimed invention is the new use of a genus of known compounds or compositions? Teva Pharm. Int’l GmbH v. Eli Lilly & Co., 172 F.4th 1367 (Fed. Cir. 2026), and In re Xencor, Inc., 130 F.4th 1350 (Fed. Cir. 2025), reach different outcomes on superficially similar facts. The contrast is instructive.

Teva v. Eli Lilly

Teva sued Lilly for infringement of three “headache patents” (U.S. Patent Nos. 8,586,045, 9,884,907, and 9,884,908) directed to reducing the incidence of, or treating, headache in a human by administering a humanized anti-CGRP antagonist antibody. Anti-CGRP antagonist antibodies bind CGRP and inhibit activity associated with headache, and murine anti-CGRP antagonist antibodies can be converted into humanized forms through known humanization methods.

The specification stated that anti-CGRP antagonist antibodies were “known in the art,” cited murine examples including antibody 4901, disclosed other murine anti-CGRP antagonist antibodies, stated that anti-CGRP antagonist antibodies “may be made by any method known in the art” (See U.S. Patent No. 8,586,045, at column 27, lines 41-42), disclosed one humanized antibody named G1, and identified prior-art humanization methods.

Representative Claim

Claim 30 of the ’045 patent is representative of the three patent claim sets. Claim 30 depends on claim 17, and rewritten with claim 17, recites: “A method for reducing incidence of or treating headache in a human, comprising administering to the human an effective amount of an anti-CGRP antagonist antibody, wherein said anti-CGRP antagonist antibody is a . . . humanized monoclonal antibody.” See U.S. Patent No. 8,586,045, at claims 17 and 30.

That language did much of the analytical work. The claim was not directed to the humanized antibody genus standing alone but was instead directed to a method of reducing the incidence of, or treating, headache by administering an effective amount of a humanized anti-CGRP antagonist monoclonal antibody. The Federal Circuit treated that distinction as substantive because the use of treating headache was different from the function that characterized the antibodies, namely antagonizing CGRP.

Procedural History

Teva brought suit in the District of Massachusetts in September of 2018, alleging Lilly indirectly infringed the headache patents via its Emgality product. The dispute then intersected with earlier IPR proceedings in which Lilly had argued that anti-CGRP antagonist antibodies were well known, prior art was replete with examples, methods of making them were extensively described, and humanization was routine. At trial, a jury found willful infringement, rejected Lilly’s written-description and enablement defenses, and awarded damages. The district court nevertheless granted JMOL for Lilly, holding the asserted claims invalid under § 112. On appeal, the Federal Circuit reviewed the JMOL de novo under First Circuit law, viewed the facts and inferences in favor of the verdict, accounted for Lilly’s clear-and-convincing burden on invalidity, and reversed because a reasonable jury could have rejected Lilly’s invalidity case.

How Procedural Posture Drove the Outcome

The posture mattered because the Federal Circuit was not writing on a blank factual record. The court treated the jury as having made all verdict-supporting factual findings that substantial evidence permitted, including findings that anti-CGRP antagonist antibodies were known, that humanization was routine, and that the specification taught the claimed therapeutic use. That standard made Lilly’s JMOL victory vulnerable: unless the record compelled invalidity despite the jury’s implicit findings and Lilly’s clear-and-convincing burden, the verdict had to be reinstated.

The Inter Partes Review (IPR) history also mattered in a practical evidentiary sense. Lilly’s prior statements about the well-known status of anti-CGRP antagonist antibodies, the extensive description of methods for making them, and the routine nature of humanization helped supply the record support that distinguished Teva from cases where a patentee merely asserted that a genus was known.

Holdings on Written Description

The Federal Circuit held that JMOL of no written description was improper. The court framed the relevant line of precedent as addressing circumstances “where a claim pertains to a well-known genus that is not, itself, the invention.” Teva, 172 F.4th at 1374. The headache patents made clear that the claimed invention was the use of anti-CGRP antagonist antibodies (or humanized versions thereof) to treat headache, not the antibodies themselves.

The court reasoned that a reasonable jury could have found the following. First, anti-CGRP antagonist antibodies and methods of making them were well known, replete, and extensively described in the prior art, based in part on Lilly’s own statements in earlier IPR proceedings. Second, humanization was a well-established and routine procedure by the November 2006 priority date. Third, a skilled artisan reading the specification would have understood that all humanized anti-CGRP antagonist antibodies treat headache.

Together with the specification’s disclosures, these findings were sufficient to support the jury’s rejection of Lilly’s written-description challenge.

The court rejected Lilly’s argument that the specification disclosed only one humanized version (G1). It reasoned that murine versions were well known, several were disclosed, and they were “just a routine (and likewise disclosed) procedure away from being humanized.” The court also rejected Lilly’s “semantic distinction without a difference” argument by distinguishing prior method-claim decisions in which the claims depended on undisclosed or unknown compounds that performed the claimed function. Here, by contrast, the claims used known anti-CGRP antagonist antibodies for the different therapeutic purpose of treating headache.

Prior Method-Claim Decisions Distinguished by the Court

The main method-claim decisions the Teva court distinguished were Univ. of Rochester v. G.D. Searle & Co., 358 F.3d 916 (Fed. Cir. 2004), and Ariad Pharms., Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010), and, in the enablement analysis, Idenix Pharms. LLC v. Gilead Scis., Inc., 941 F.3d 1149 (Fed. Cir. 2019). As characterized by Teva, University of Rochester involved method claims directed to selectively inhibiting PGHS-2 by administering a compound that selectively inhibits PGHS-2. The Teva court explained that the method-versus-compound distinction did not help the Rochester patentee because the specification disclosed no compounds usable in the claimed methods, and there was no evidence that such compounds were known.

As characterized by Teva, Ariad involved method claims directed to reducing NF-ÎşB activity using substances that achieved that reduction. The Teva court described the problem as insufficient disclosure of the molecules capable of performing the claimed reduction, including hypothesized categories with no adequate examples or insufficient evidence of the required activity.

The distinction matters because Teva viewed those cases as claims to methods of doing X using something that does X. By contrast, the headache claims were not framed as methods of antagonizing CGRP using antibodies that antagonize CGRP. They were methods of treating headache using known anti-CGRP antagonist antibodies, which the court treated as a different invention from the function that characterized the antibodies.

In the enablement discussion, the court also distinguished Idenix, which Teva characterized as involving a method of treating hepatitis C with a large class of 2’-methyl-up nucleosides. The key difference was evidentiary: in Idenix, the record did not support a finding that all claimed nucleosides treated HCV, while in Teva a reasonable jury could find that all humanized anti-CGRP antagonist antibodies treated headache.

Holdings on Enablement

The Federal Circuit noted that “enablement and written description often rise and fall together” on facts like these. Ariad, 598 F.3d at 1352. The court held that JMOL of no enablement was also improper because the relevant enablement question followed the claim language. Lilly’s breadth argument assumed that the patents had to enable the full antibody genus as such, but the asserted claims covered “only the use of such antibodies for the different, limited purpose of treating headache.” Teva, 172 F.4th at 1381. Because the record supported a finding that all humanized anti-CGRP antagonist antibodies treat headache, the court treated the task of finding or making more antibodies as “more akin to extra credit than a necessary research assignment.” Id. That is a direct holding of Teva, but it is also a posture-sensitive holding because it depended on what the jury could reasonably find from the specification, prior art, and Lilly’s admissions.

In re Xencor

Facts and Technology

The application at issue related to modified antibodies with Fc substitutions M428L/N434S that increased in vivo half-life and were intended to reduce the need for frequent treatments. Anti-C5 antibodies bind the C5 complement protein, and the specification identified one anti-C5 antibody, 5G1.1. Eculizumab, a humanized version of 5G1.1 with a known sequence, had been used to treat PNH and studied for other indications. The prosecution problem was that this technical disclosure did not translate into written-description support for the full treatment language Xencor chose.

Claims at Issue

Claims 8 and 9 are representative of the claims at issue.

Claim 8, a Jepson Claim, recites:

“[i]n a method of treating a patient by administering an anti-C5 antibody with an Fc domain, the improvement comprising said Fc domain comprising amino acid substitutions M428L/N434S . . . wherein said anti-C5 antibody with said amino acid substitutions has increased in vivo half-life as compared to said antibody without said substitutions.” Xencor, 130 F.4th at 1354-55.

Claim 9 is a method claim with a means-plus-function element, and it recites:

“[a] method of treating a patient by administering an anti-C5 antibody comprising: a) means for binding human C5 protein; and b) an Fc domain comprising amino acid substitutions M428L/N434S . . . wherein said anti-C5 antibody with said amino acid substitutions has increased in vivo half-life as compared to said antibody without said substitutions.” Xencor, 130 F.4th at 1355.

The consequential words were not limited to the Fc substitutions. For claim 8, the Jepson preamble, which recited “[i]n a method of treating a patient by administering an anti-C5 antibody with an Fc domain,” became part of the claimed invention. For claim 9, the phrase “[a] method of treating a patient by administering an anti-C5 antibody” was limiting because it gave meaning to administering, to the claimed in vivo half-life improvement, and to the therapeutic context of the method. The court therefore analyzed written description against the full treatment language, not merely the Fc-engineering improvement.

Procedural History

The examiner rejected claims 8 and 9 for lack of written description in 2021. The Board affirmed in January 2023, Xencor sought rehearing, and Xencor appealed to the Federal Circuit. Before that appeal was heard, the United States Patent and Trademark Office requested remand for ARP review, and the ARP issued a superseding decision on May 21, 2024, maintaining the written-description rejections. The Federal Circuit then affirmed on March 13, 2025, reviewing legal issues de novo, reviewing Board fact findings for substantial evidence, and deferring to the Board’s expert-credibility determinations. That procedural posture was decisive: unlike Teva, where the appellate court protected jury-supported findings against JMOL, Xencor came to the court with adverse agency fact findings that were reviewed deferentially.

Holdings on Written Description

The Federal Circuit held that “treating a patient” was limiting because the phrase was directly connected to “administering an anti-C5 antibody” by the word “by,” the claim lacked dosage and rate, and “increased in vivo half-life” made practical sense only in a treatment context. The specification reinforced that reading by explaining that longer in vivo half-lives allow less frequent injections or lower dosing for therapeutic antibodies. The result was that “treating a patient” was not a decorative statement of purpose but rather supplied the raison d’ĂŞtre of the method and expanded the written-description burden.

The Federal Circuit also held that the preamble of a Jepson claim requires written description because Jepson form uses the preamble to define the claimed invention and limit claim scope. Claim 8 therefore could not be reduced to the improvement alone, namely “said Fc domain comprising amino acid substitutions M428L/N434S.” The invention was the improvement as applied to the recited prior-art method of treating a patient by administering an anti-C5 antibody, and Xencor could not avoid written description by simply asserting that the preamble subject matter was already known.

Because the specification did not limit treatment to a specific disease, the ARP and Federal Circuit treated “treating a patient” as extending to all patients and all diseases. The specification did not define “treating,” did not provide data associated with treating any patient with any disease or condition using any anti-C5 antibody with the claimed Fc modifications, and did not include a working example of treatment. Listing three general disease or condition classes and unmodified antibodies was therefore inadequate to show possession of the claimed treatment method across the full scope of the claim.

The ARP found the single disclosed anti-C5 antibody, 5G1.1, insufficient for the broader anti-C5 antibody genus because the genus involved different specificities and epitopes. It also adopted the Board’s findings that Xencor’s expert was not credible and that the prior-art examples did not establish that anti-C5 antibodies were so well known that the specification could omit further support. The Federal Circuit affirmed because those adverse factual findings were supported by substantial evidence.

Section 112(f) and MPEP Section 2181

Claim 9’s “means for binding human C5 protein” limitation invoked 35 U.S.C. § 112(f) (formerly § 112, ¶ 6). Under MPEP § 2181, a limitation using “means” or “step” or another nonce term, reciting function, and failing to recite sufficient structure is construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof. The ARP identified 5G1.1 as the sole corresponding structure and found that disclosure adequate for the means-plus-function limitation itself. A person of ordinary skill would understand 5G1.1 to refer to the original mouse monoclonal antibody and eculizumab, which is a humanized version. Equivalents of 5G1.1 need not be separately described.

However, claim 9 still lacked written description for the separate “treating a patient” limitation. The § 112(f) analysis narrowed the antibody element to 5G1.1 and equivalents, but that narrowing did not supply written description for the broad, unlimited treatment preamble. MPEP § 2181 is claim-interpretation guidance that anchors claim scope to disclosed structures and equivalents, but it does not resolve § 112(a) adequacy for other claim limitations.

Comparing Teva and Xencor

Factual Records

In Teva, the factual record was robust. Lilly’s own IPR statements admitted that anti-CGRP antagonist antibodies were “well known,” that prior art was “replete” with such antibodies, that methods of making them were “extensively described,” and that humanization was “a well-established and routine procedure.” Expert testimony and record evidence corroborated these points, and the Federal Circuit held a reasonable jury could credit that evidence.

In Xencor, no comparable record existed. The Board found Xencor’s expert not credible, the prior-art examples insufficient to establish that anti-C5 antibodies were well-known, and the specification limited to a single anti-C5 example, 5G1.1.

Scope of the Specification Relative to Claim Breadth

In Teva, the claims were directed to “reducing incidence of or treating headache” with humanized anti-CGRP antagonist antibodies, and the specification’s disclosures were evaluated in light of record evidence that the antibody genus and humanization methods were already known. In Xencor, the operative phrase was the much broader “treating a patient,” which the court treated as untethered to any particular disease, patient population, dose, or outcome. The specification provided no treatment data, no working example of treating any disease with a modified anti-C5 antibody, and no definition of “treating.” The mismatch between claim scope and specification content did not help Xencor’s case.

Role of POSA Knowledge

Both decisions acknowledge that POSA (person of ordinary skill in the art) knowledge supplements specification disclosures. Teva demonstrates that when POSA knowledge is established in the record through admissions, expert testimony, and documentary evidence, it can bridge gaps between what the specification explicitly discloses and what the claims cover. Xencor demonstrates that POSA knowledge cannot be invoked by bare assertion. The applicant must show that what is claimed as well-known is, in fact, supported by the record.

Application of Enablement and Written Description to Method Claims

In Teva, written description and enablement were addressed together, and the Federal Circuit noted that they “often rise and fall together.” The key was that the claimed use was the limited purpose of treating headache, not ownership of the entire antibody genus for all purposes. In Xencor, only written description was directly decided by the Federal Circuit. The breadth of the treatment limitation, combined with the absence of treatment data or examples, defeated written description without requiring the court to reach enablement.

Counterfactual: Xencor Under a Teva v. Eli Lilly Claim-Construction Lens

It is unlikely that Xencor would have fared materially better under a Teva-style focus on subsidiary factual determinations in claim construction. The Federal Circuit in Xencor already separated legal claim-construction questions from factual written-description findings: it reviewed legal determinations de novo, reviewed Board fact findings for substantial evidence, and deferred to the Board’s expert-credibility determinations. That posture left Xencor with two adverse fact-finding problems that a more fact-sensitive claim-construction approach would not cure. The Board did not credit Xencor’s expert on whether anti-C5 antibodies were well known, and the Board found the prior-art examples insufficient to establish the genus as well known.

The intrinsic record did help Xencor in one limited respect: it supported construing the claims as therapeutic methods. The specification linked antibody administration to treatment frequency, clearance, half-life, patient characteristics, and therapeutic use. But that same intrinsic evidence cut against Xencor once “treating a patient” was held limiting, because it confirmed that the claimed method needed written-description support as a treatment method, not merely as an Fc-engineering platform.

However, the specification was too sparse to survive that burden. It did not identify what patients with what diseases or conditions could be treated with an anti-C5 antibody carrying the claimed Fc substitutions, did not include a working example of treating any disease with such an antibody, and did not provide data associated with treating any patient using the claimed modified anti-C5 antibodies. Even claim 9’s § 112(f) narrowing of the antibody element to 5G1.1 and equivalents did not fill the separate gap in treatment disclosure.

A stronger Xencor claim might have looked more like the claim in Teva: a specific disease, a specific antibody or supported antibody class, and a specification explaining why the claimed class works for that disease. On the record actually before the court, however, Xencor had broad treatment language, a single anti-C5 example, no treatment example, and no credited showing that the anti-C5 genus was well known. Under any fact-sensitive construction, the problem was not simply how the preamble was read but rather that the specification did not support the breadth that the preamble created.

The Central Legal Issue: Method Claims Using Known Compounds

As such, the central question is: when a method-of-use claim covers the use of known compounds or compositions, and the invention is the new use rather than the molecule itself, does the specification satisfy enablement and written description requirements even if it does not exhaustively describe or reduce to practice every member of the claimed composition genus?

Teva supports the proposition that a method-of-use claim may survive written description and enablement challenges under these circumstances if certain conditions are met. The composition genus must be well-known in the art and must not itself be the invention. The specification and record must show that a person of ordinary skill in the art would know how to make or obtain the genus components. The specification must convey that the entire claimed genus works for the specific claimed method, such as the finding in Teva that all humanized anti-CGRP antagonist antibodies treat headache. Finally, the POSA knowledge must be established in the evidentiary record rather than merely asserted.

This is not a universal rule. The Teva court expressly noted that if the claims had covered the antibody genus itself, the analysis would differ.

Xencor limits this principle in at least four important ways. First, POSA knowledge cannot be invoked by assertion but must be established in the record through credible evidence. Second, broad treatment language such as “treating a patient” without disease limitation can create a treatment genus so large that no reasonable specification disclosure can support it. Third, a section 112(f) limitation may narrow a structural component to known or disclosed corresponding structure and equivalents, but it will not supply written description for a separate, broad treatment limitation. Fourth, Jepson claiming does not eliminate the written description requirement for the preamble and indeed makes the preamble affirmatively limiting.

The distinction between the two outcomes ultimately reduces to specificity, evidentiary support, and standard of review. The claims in Teva targeted a specific therapeutic purpose, namely reducing incidence of or treating headache, and the appellate posture required the Federal Circuit to credit jury-supported findings about POSA knowledge and therapeutic applicability. The claims in Xencor targeted undifferentiated treatment of a patient, and the appellate posture required deference to adverse agency fact findings that the anti-C5 genus was not shown to be well known and that the specification did not support the full treatment scope.

One of the genuine privileges of practicing as a patent attorney is witnessing technologies evolve from concept to real-world application, particularly when they address long-felt, unmet needs. The announcement this week from UC San Diego offers a compelling example of exactly that trajectory.

UC San Diego recently announced the formal launch of its . The new institute is a multidisciplinary hub designed to leverage AI and digital technology to improve health care delivery and patient outcomes. As the university described it, the institute will “harness transformative ideas, backed by rigorous research and implemented with agility, to achieve scalable, reliable health outcomes for all.” The institute unifies experts from UC San Diego Health and faculty across six academic schools. These include the School of Medicine, the Jacobs School of Engineering, the Skaggs School of Pharmacy and Pharmaceutical Sciences, the Rady School of Management, the HalıcıoÄźlu School of Data Science and Computing, and the Herbert Wertheim School of Public Health and Human Longevity Science. It will also draw on specialized resources including the San Diego Supercomputer Center and the School of Medicine’s Division of Biomedical Informatics. 

The institute is built on three strategic pillars: health intelligence and implementation leadership, training and education, and innovation, outcomes, and research. Its goal is to unite more than 50 faculty and empower 750 trainees. 

This development resonates with themes I explored in a prior post on AI and drug repurposing. That piece examined a review article by Fu et al. published in the Annual Review of Medicine, which highlighted how the explosion of multiomics datasets, including genomics, transcriptomics, proteomics, metabolomics, and radiomics, combined with increasingly digitized electronic health records, presents both an extraordinary opportunity and a formidable challenge for AI-driven pharmaceutical discovery. The common thread is clear: digital technology and machine learning are no longer theoretical tools at the periphery of health care; they are being embedded into the institutional infrastructure of research and clinical practice.

From my vantage point as an IP attorney, I am watching this space closely. The application of AI and machine learning to improve patient health, whether through drug repurposing, predictive analytics, or the kind of large-scale digital transformation UC San Diego envisions, represents one of the most important innovation trends of our time. As these technologies mature, so too will the intellectual property questions surrounding them. I look forward to continuing to track these developments and sharing observations as they unfold.

On June 30, 2026, the National Institutes of Health announced that its All of Us Research Program has become the world’s largest integrated genomic and electronic health record database, with data from more than 747,000 participants now available to researchers.  (NIH). This is more than a programmatic milestone. It represents a direct answer to one of the most persistent barriers facing AI in drug repurposing: the lack of large, diverse, and integrated datasets that combine genomic depth with real clinical data.

This problem was highlighted in my recent post reviewing the use of AI and machine learning in drug repurposing, drawing on the work of Fu et al. in the . AI and Drug Repurposing: Old Drugs, New Tricks, and the IP Questions (Post). One of the key challenges I highlighted is that existing multiomics and clinical data arise from heterogeneous patient samples across different laboratories and health care systems, making harmonization extremely difficult. AI at 391. Limited data sharing between biopharmaceutical companies and academic institutions compounds the problem, as proprietary datasets cannot be accessed by the broader research community due to intellectual property concerns. Post at 2. The result is that the growing mass of genetic and multiomics data has not been effectively explored for drug repurposing due to a lack of accurate, integrated approaches. AI at 382.

The All of Us release speaks directly to these barriers. The dataset now includes more than 535,000 whole genome sequences linked to nearly 482,000 electronic health records, a combination of genomic depth and clinical breadth unmatched by any other research program in the world. NIH at 1. It encompasses more than 1.3 billion genetic variants, 553,000 genotyping arrays, and 96,000 structural variant records, alongside 747,000 survey responses capturing social circumstances, behaviors, and environments. Id. For the first time, the dataset also includes proteomics and RNA sequencing data, moving the program into the multiomics era. NIH at 2. 

Two additional features make this resource particularly relevant for AI. More than 86% of participants come from communities historically underrepresented in biomedical research, spanning all 50 states and more than 98% of U.S. three-digit ZIP codes. NIH at 1-2. That diversity addresses another concern raised in the article: that real world data is challenged by confounding factors such as sex, race, and socioeconomic status, as well as a lack of detailed clinical, biomarker, and genetic information. AI at 387. All of Us data is available to registered researchers at no cost, giving scientists at rural universities the same access as those at major research institutions. NIH at 2. 

The program has already fueled more than 1,400 peer-reviewed publications, including work identifying existing medications that may help prevent Alzheimer’s disease. Id. As NIH Director Jay Bhattacharya noted, “To tailor treatments to individuals, you actually need very large populations to uncover the patterns that connect genetics, lifestyle, and the environment to health outcomes.” NIH at 1.

That observation captures why this matters for AI and drug repurposing. The field has algorithms and computational power. A unified, diverse, openly accessible data foundation connecting genomes to clinical realities is a missing critical piece. 

“To tailor treatments to individuals, you actually need very large populations to uncover the patterns that connect genetics, lifestyle, and the environment to health outcomes.”

-NIH Director Jay Bhattacharya

A fascinating review article recently published in the Annual Review of Medicine by Fu et al. caught my attention for what it signals about the future intersection of artificial intelligence and pharmaceutical development, and the intellectual property questions that may follow. “” (Fu et al., Annu. Rev. Med. 2026; 77:381–398), provides a comprehensive survey of how AI and machine learning (ML) are being deployed to find new therapeutic uses for known drugs. 

The Data Explosion Driving Discovery

The premise is straightforward — we are awash in biological data. The rapid growth of multiomics datasets (genomics, transcriptomics, proteomics, metabolomics, and radiomics) together with increasingly digitized electronic health records, presents both an extraordinary opportunity and formidable challenge. The sheer volume of available data has outpaced researchers’ ability to explore it using traditional methods, and that gap is precisely where AI and ML tools are stepping in. 

Drug repurposing seeks to identify new indications for known drugs whose pharmacokinetics, safety profiles, and manufacturing processes may be already well-characterized and is promised to offer faster and less expensive paths to treatment than de novo drug development. 

Two Frameworks for Repurposing Known Drugs

Fu et al. describe two major strategic approaches. The first is target-centric drug repurposing, which rests on the premise that the same target protein may be implicated in multiple diseases; a drug acting on that target could therefore treat conditions beyond its original indication. Fu, et al. at  382. For example, Metformin is being explored for atrial fibrillation, and semaglutide, the GLP-1 receptor agonist, is being investigated for neurodegenerative diseases and substance abuse in addition to its approved diabetes and obesity indications. Id. at p. 384.

The second approach is disease-centric drug repurposing, which identifies drugs that may work across diseases sharing similar biological pathways, symptoms, or traits. A key step here is identifying overlapping mechanisms of action between the original and target diseases. The well-known example of sildenafil, originally for erectile dysfunction and pulmonary hypertension, now investigated for Alzheimer’s disease, illustrates how network-based analysis of disease modules can surface unexpected therapeutic candidates. Id. at 382.

The AI Toolkit

The computational infrastructure behind AI-driven repurposing draws on five categories of databases: chemoinformatic, bioinformatics, systems biology, multiomics, and pharmacological databases. Id. at 384-388. These provide the “AI-ready” datasets on which models are trained and evaluated.

AI subsets of machine learning and deep learning use algorithms to learn patterns from input data, and are categorized into supervised learning (labeled data required), unsupervised learning (no labels), and semi-supervised learning (partially labeled). Id. at 386. Network-based approaches, which analyze graph data representing relationships among drugs, targets, genes, and pathways, are particularly powerful in disease-centric repurposing, excelling at uncovering previously undiscovered relationships between drugs and their potential targets. Id. at 387.

Computational predictions, however promising, must be validated. The authors emphasize that experimental and clinical validation are crucial steps following AI-based repurposing, as they confirm real-world accuracy and reliability. Id. at 389. Validation approaches include iPSC-derived models (e.g., neurons from Alzheimer’s patients), animal models, and real-world evidence drawn from electronic health records or health insurance claims data. Id. at 383.

Demonstrated Success and Persistent Challenges

AI-driven repurposing has already yielded meaningful results for the treatment of Alzheimer’s disease, cardiovascular disease, cancer and COVID-19. Id. at 389-390.

Yet the challenges to widespread adoption are significant. Complex AI models demand computationally expensive hyperparameter tuning. Id. at 391. Large language models, while promising for integrating heterogeneous data, are susceptible to hallucinations and biased outputs. Id. Data security remains a concern: patient-level multiomics and clinical data require robust protections such as federated learning and interoperability standards. Id. Perhaps most critically, multiomics and clinical data arise from heterogeneous patient samples across different laboratories and health care systems, making harmonization difficult. Id. Progress will require genuine cross-disciplinary collaboration between biopharmaceutical companies, academic institutions, clinicians, and computational scientists working together. Id. 

The IP Angle: Why Patent Attorneys Should Be Paying Attention

From an intellectual property standpoint, AI-driven drug repurposing raises questions worth tracking. When an algorithm identifies a new use for an existing compound, what is the patentable invention: the method of treatment, the algorithm itself, or the data pipeline that made discovery possible? Can novel analogs and derivatives be developed? Does the repurposing require new formulations and dosing schedules? Second, medical-use claims have long been part of the pharmaceutical patent toolkit, but the involvement of AI in generating those discoveries may complicate inventorship analysis and raise questions about the sufficiency of disclosure and enablement.

There is also the question of data as a competitive asset. Fu et al. note that biopharmaceutical companies possess massive datasets that cannot be shared due to intellectual property concerns. The proprietary databases, curated knowledge graphs, and trained models underlying AI-driven repurposing represent significant intangible value potentially protectable through trade secret, patent, or a combination of strategies.

For those of us advising clients in this space, the conclusion of the Fu et al. review resonates: “AI is an indispensable aspect of the future of drug repurposing and precision medicine for challenging human diseases”.  Id. at 391. The legal frameworks surrounding these discoveries will need to evolve in tandem with science.

Neural tube defects are among the most challenging congenital conditions in medicine, with a devastating burden on patients, families, and health care systems. These defects arise when the neural tube, the embryonic structure that becomes the brain and spinal cord, fails to close properly during early fetal development. This failure can range in severity from minor disabilities to lethal malformations. Globally, nearly half a million babies are born each year with such defects. Spina bifida, one of the most recognized forms, continues to occur despite preventive measures such as folate supplementation. In the United States, annual hospital charges related to spina bifida were estimated at two billion dollars in 2024, underscoring the economic weight of lifelong multidisciplinary care for affected children.

Myelomeningocele is the most severe form of spina bifida. In this condition, the spinal cord and surrounding tissues remain exposed due to failure of closure, leading to leakage of cerebrospinal fluid. The developing spinal cord is then subjected to chemical damage from amniotic fluid and mechanical trauma from contact with the uterine wall. This dual insult results in a host of developmental abnormalities. Patients may suffer loss of motor function below the lesion level, causing paralysis, as well as bowel and bladder dysfunction. A common associated problem is hindbrain herniation, in which the lower portions of the brain — specifically the cerebellum and brainstem — are pulled downward from their normal position into the upper spinal canal. This can disrupt normal brain and spinal fluid flow, cause pressure on delicate brain structures, and lead to neurological symptoms.

The newly reported phase 1 trial explored a groundbreaking regenerative approach to myelomeningocele repair before birth — “.” The therapy uses placenta-derived mesenchymal stem cells, known as PMSCs, obtained from early-gestation placentas. These cells are uniquely suited for fetal applications due to their fetal origin, lack of tumor-forming potential, neuroprotective properties, and ability to reduce cell death (apoptosis). Importantly, when cultured in a specialized neurogenic medium, PMSCs adapt to secrete higher concentrations of neurotrophic growth factors and can rescue dying neurons in laboratory settings.

In this trial, PMSCs were seeded onto an FDA-approved extracellular matrix graft and placed directly onto the exposed fetal spinal cord during standard prenatal surgery for myelomeningocele. The aim was to augment the mechanical effects of surgical closure with a biologically active therapy capable of protecting and regenerating neural tissue during a crucial window of fetal neurodevelopment. Investigators designed the study to rigorously monitor surgical feasibility and early safety, given the unknowns of placing living allogeneic stem cells into the developing central nervous system.

The results are highly encouraging. All six patients in the initial cohort successfully received the therapy, and no stem cell-related adverse events were observed. Every newborn had reversal of hindbrain herniation on postnatal MRI, with intact and fully healed surgical repair sites. There was no evidence of abnormal tissue growth or tumor formation. The therapy did not interfere with the known benefits of fetal surgery, nor did it cause complications in wound healing. These findings suggest that PMSCs can be safely integrated into prenatal surgery without compromising established surgical gains.

This pioneering work also highlights the growing intersection between regenerative medicine and personalized medicine. Personalized medicine is not only about genetic diagnostics and targeted drugs, it increasingly encompasses designing and delivering the right biologic or regenerative therapy at the right developmental moment for maximum impact. In the case of myelomeningocele, early, targeted intervention with a biologically active repair tailored to the unique characteristics of the patient’s lesion could represent the next evolution in precision perinatal care.

The implications go beyond this single condition. This trial demonstrates a scalable and clinically feasible platform for delivering biologically active therapeutics directly to the fetus. By intervening early in development, such therapies may alter lifelong health trajectories, potentially preventing paralysis, preserving bladder and bowel function, and reducing the societal and economic impact of chronic disability. While longer-term follow-up is needed to confirm sustained benefits, this first-in-human study provides hope for addressing a serious unmet medical need for which no curative alternatives exist. It also adds to the very short list of regenerative medicine therapies reaching clinical application, particularly in the prenatal setting.

In January 2026, the Federal Circuit issued a nonprecedential opinion in Guardant Health, Inc. v. University of Washington (Slip Op. 2024-1129, Jan. 23, 2026) that, while not binding precedent, is nevertheless highly relevant to patent practice in any technology. The patent at issue, U.S. Patent No. 10,760,127, claims methods for reducing sequencing errors using Duplex Consensus Sequencing (DCS), a technique that pairs double-stranded DNA with unique molecular identifiers before amplification and sequencing. This approach is designed to achieve unprecedented sensitivity in detecting rare variants, a capability critical to applications like early cancer detection and other high-stakes medical diagnostics.

Guardant had challenged the patent before the Patent Trial and Appeal Board (PTAB), arguing that the claims were obvious under 35 U.S.C. § 103, over a combination of references. The elements at issue were:

b) amplifying original strands of at least a portion of the double-stranded adapter-DNA molecules to produce first and second strand copies;

c) sequencing a plurality of first and second strand copies to obtain first and second strand sequence reads for at least a portion of the adapter-DNA molecules;

Slip Op. at 4.

Guardant argued that these two claim elements were disclosed together in a single embodiment in a prior art reference, U.S. Patent Application No. 2009/0298075 A1[1] (Travers ’075). Travers ’075, while primarily directed to PacBio’s SMRT sequencing, expressly described in a single paragraph performing rolling circle replication (RCR) to produce repeated sense and antisense strands, followed by sequencing those amplified products. Washington countered with Travers 2010[2], another publication from PacBio scientists. Travers 2010 described the SMRTbell template, a topologically circular DNA structure with a double-stranded region and single-stranded hairpin loops, which can be sequenced using SMRT technology without any amplification step. In Washington’s view, the general teaching in Travers 2010 that SMRT sequencing does not depend on amplification reinforced their argument that a skilled artisan would need additional motivation to incorporate RCR amplification before SMRT sequencing and might doubt the feasibility of such a combination.

The PTAB agreed with Washington, concluding Guardant had not shown a motivation to combine the amplification concept from Travers ’075 with the sequencing process of SMRTbell templates as taught in Travers 2010, nor a reasonable expectation of success in doing so.[3] But the Federal Circuit disagreed, explaining that when claim elements appear together in one embodiment of a single reference, here, amplification and sequencing in Travers ’075, no separate “motivation to combine”[4] or “reasonable expectation of success”[5] analysis is required as if the elements originated from disparate sources. Because Guardant’s petition relied on a single  paragraph of Travers ’075 itself, and not on modifying the SMRT sequencing process described in Travers 2010, the PTAB’s approach was legally erroneous. The court vacated the PTAB’s ruling and remanded for consideration of whether Travers ’075’s RCR-plus-sequencing disclosure was enabled, a separate requirement for relying on it in an obviousness analysis.

Two key takeaways emerge from the opinion. First, in the context of obviousness, if all disputed claim limitations are disclosed together in one embodiment in a single prior art reference, the analysis does not require a separate finding of motivation to combine those elements. Treating co-disclosed steps as if they came from isolated references improperly burdens the challenger and can distort the obviousness inquiry. Second, even when a prior art reference makes only brief mention of a process or embodiment, in this case, rolling circle replication followed by SMRT sequencing, that disclosure remains prior art “for all that it teaches.” Whether that disclosure is enabled is a separate question that must be addressed before relying on it to establish obviousness.

Now is the perfect time to reassess ongoing patent challenges and diligence frameworks to ensure internal strategies fully reflect this clarified standard. To ensure your patent strategy is ahead of the game, contact your Â鶹ֱ˛Ą attorney for guidance and to proactively review your portfolio.

---

[1] U.S. Patent Application No. 2009/0298075 A1

[2] Travers et al., A Flexible and Efficient Template Format for Circular Consensus Sequencing and SNP Detection, 38 Nucleic Acids Research e159 (2010)

[3] Slip Op. at 8.

[4] Slip Op. at 9.

[5] Slip Op. at 13.

---