Case: 24-1078 Document: 46 Page: 1 Filed: 11/18/2025

United States Court of Appeals for the Federal Circuit ______________________

DUKE UNIVERSITY, ALLERGAN SALES, LLC, Plaintiffs-Appellees

v.

SANDOZ INC., Defendant-Appellant ______________________

2024-1078 ______________________

Appeal from the United States District Court for the District of Colorado in No. 1:18-cv-00997-RM-KLM, Judge Raymond P. Moore. ______________________

Decided: November 18, 2025 ______________________

JEFFREY A. LAMKEN, MoloLamken LLP, Washington, DC, argued for plaintiffs-appellees. Also represented by KAYVON GHAYOUMI, MICHAEL GREGORY PATTILLO, JR.; ELIZABETH KATHLEEN CLARKE, Chicago, IL; SARA MARGOLIS, New York, NY; LISA BARONS PENSABENE, HASSEN A. SAYEED, O'Melveny & Myers LLP, New York, NY.

WILLIAM M. JAY, Goodwin Procter LLP, Washington, DC, argued for defendant-appellant. Also represented by GABRIEL FERRANTE, New York, NY; VISHAL C. GUPTA, Step- toe LLP, New York, NY; ROBERT KAPPERS, Chicago, IL. Case: 24-1078 Document: 46 Page: 2 Filed: 11/18/2025

2 DUKE UNIVERSITY v. SANDOZ INC.

______________________

Before DYK, STOLL, and STARK, Circuit Judges. STARK, Circuit Judge. Sandoz Inc. (“Sandoz”) appeals from a judgment of the U.S. District Court for the District of Colorado, holding that Sandoz failed to prove claim 30 of U.S Patent No. 9,579,270 (the “’270 patent”) invalid for lack of ade- quate written description. We reverse. I Plaintiffs-appellees Duke University and Allergan Sales, LLC (collectively referred to herein as “Allergan”) to- gether own all rights in the ’270 patent and certain related patents. The ’270 patent issued in 2017 and has a priority date of 2000. It is entitled, “Compositions and Methods for Treating Hair Loss Using Non-Naturally Occurring Pros- taglandins,” and relates generally to treating hair loss us- ing compositions containing prostaglandin F (“PGF”) analogs. ’270 pat. abstract. As the district court explained: The ’270 Patent describes a method for growing hair by topically applying a chemical compound known as a prostaglandin. Prostaglandins are mol- ecules that bind to certain receptors on cells in a living body and change how such cells function. The human body produces a variety of prostaglan- dins; the general type at issue here is known as prostaglandin F or PGF. Within the general cate- gory of PGF are many variants, some naturally-oc- curring and some synthesized. These variants are referred to as PGF analogs. Analogs differ from one another by virtue of various molecules that can attach to [the] base structure of the prostaglandin and which change its pharmacological properties. For example, the prostaglandin is much like a Case: 24-1078 Document: 46 Page: 3 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 3

charm bracelet to which different charms can be at- tached at different points. J.A. 5-6. At issue in this appeal is claim 30 of the ’270 patent. When read together with claims 17, 24, and 25, from which it depends, claim 30 recites: A method of growing hair, wherein the method comprises topically applying to mammalian skin a safe and effective amount of a composition compris- ing: . . . an active ingredient selected from the group consisting of a prostaglandin F analog of the follow- ing structure:

and pharmaceutically acceptable salts thereof; wherein R1 is C(O)NHR3 [i.e., an amide]; R2 is a hydrogen atom; R3 is methyl, ethyl, or isopropyl; X is selected from the group consisting of —C≡C—, a covalent bond, —CH=C=CH—, —CH=CH—, —CH=N—, —C(O)—, Case: 24-1078 Document: 46 Page: 4 Filed: 11/18/2025

4 DUKE UNIVERSITY v. SANDOZ INC.

—C(O)Y—, and —(CH2)n—, wherein n is 2 to 4; Y is selected from the group consisting of a sulfur atom, an oxygen atom, and NH; and Z is phenyl. ’270 pat. at 66:65-67:54. Allergan markets Latisse®, an FDA-approved topical solution for treatment of eyelash hair loss by stimulating hair growth. J.A. 10078-79. Latisse® consists of a 0.03% bimatoprost ophthalmic solution. J.A. 16. Bimatoprost is a PGF analog with an ethyl amide at its action end (the “C1 end” or “C1 location” and also known as the “R1” position), and phenyl at the omega end, which is also referred to as the “Z” position. 1 Sandoz manufactures and sells a generic version of La- tisse®. In 2018, Allergan sued Sandoz, alleging its generic drug product infringes claim 30 of the ’270 patent. J.A. 77. Sandoz stipulated to infringement but challenged the va- lidity of the claim. J.A. 1-2, 7783. During a five-day jury trial, Sandoz attempted to prove that claim 30 of the ’270 patent is invalid for lack of ade- quate written description. In support of its contention, Sandoz presented the testimony of its expert, Dr. Clayton Heathcock. Dr. Heathcock opined that claim 30 lacks suf- ficient written description because “the Claim describes over 4,000 compounds that can cause hair to grow” but does not identify “a single” specific embodiment of the claim in the specification or disclose sufficient common structural features of the compounds encompassed by the claim. J.A. 10448-56. Allergan countered Sandoz’s evidence with

1 Like the parties, we use C1 position and R1 inter-

changeably throughout this opinion. We do the same for omega end and Z. Case: 24-1078 Document: 46 Page: 5 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 5

testimony from its own expert, Dr. Allen Reitz, who opined that the ’270 patent “adequately describes the use of am- ides for growing hair . . . with three types of prostamides with a phenyl ring at the end of the omega chain.” J.A 10679. The jury found Sandoz had failed to prove that claim 30 was invalid for obviousness, lack of enablement, or lack of adequate written description and awarded Allergan $39 million in infringement damages. J.A. 8514-15. Sandoz filed a motion for a new trial and for judgment as a matter of law, both of which the district court denied. J.A. 37-50. Sandoz timely appealed. The district court had jurisdiction under 28 U.S.C. § 1338(a), and we have jurisdiction pursu- ant to 28 U.S.C. § 1295(a). II When a district court denies a motion for judgment as a matter of law, we review its decision under the law of that court’s regional circuit. See Promega Corp. v. Life Techs. Corp., 875 F.3d 651, 659 (Fed. Cir. 2017). This case arises out of the Tenth Circuit, which applies de novo review to a district court’s denial of a motion for judgment as a matter of law. See Stroup v. United Airlines, Inc., 26 F.4th 1147, 1156 (10th Cir. 2022). Thus, “a district court’s refusal to grant judgment as a matter of law may be reversed only if the evidence is such that without weighing the credibility of the witnesses the only reasonable conclusion is in the moving party’s favor.” Id. (quoting Elm Ridge Expl. Co., LLC v. Engle, 721 F.3d 1199, 1216 (10th Cir. 2013) (inter- nal brackets and quotation marks omitted). Whether a patent provides adequate written descrip- tion for a claim presents a question of fact. See Gen. Hosp. Corp. v. Sienna Biopharmaceuticals, Inc., 888 F.3d 1368, 1371 (Fed. Cir. 2018). That question is whether the patent specification discloses that a person of ordinary skill in the art would conclude “the inventor possesse[d] the full scope of the invention” at the time of their patent application. Case: 24-1078 Document: 46 Page: 6 Filed: 11/18/2025

6 DUKE UNIVERSITY v. SANDOZ INC.

LizardTech, Inc. v. Earth Res. Mapping, Inc., 424 F.3d 1336, 1345 (Fed. Cir. 2005); see also 35 U.S.C. § 112(a) (“The specification shall contain a written description of the invention . . . .”); Carnegie Mellon Univ. v. Hoffmann- La Roche Inc., 541 F.3d 1115, 1124 (Fed. Cir. 2008) (“[T]o satisfy the written description requirement for a claimed genus, a specification must describe the claimed invention in such a way that a person of skill in the art would under- stand that the genus that is being claimed has been in- vented, not just a species of the genus.”). Like all invalidity defenses, lack of written description must be proven by clear and convincing evidence. See Vasudevan Software, Inc. v. MicroStrategy, Inc., 782 F.3d 671, 682 (Fed. Cir. 2015). Thus, Sandoz’s “burden on appeal is doubly high,” as it “must show that no reasonable jury could have failed to conclude that [its] case had been established by clear and convincing evidence.” Boehringer Ingelheim Vetmedica, Inc. v. Scherling-Plough Corp., 320 F.3d 1339, 1353 (Fed. Cir. 2003). III While Sandoz raises multiple issues on appeal, 2 we need address only one: its contention that the district court should have granted judgment as a matter of law that claim 30 of the ’270 patent is invalid for lack of adequate written description. Sandoz argues that claim 30 covers the use of a specific subgenus of PGF analogs to grow hair, but that the specification fails to show a skilled artisan that

2 Sandoz’s other principal arguments on appeal are that Allergan should have been collaterally estopped from litigating its obviousness defense due to prior decisions of this court, that the district court’s instruction to the jury on teaching away was prejudicially flawed, and that claim 30 was not enabled. Because we have determined that claim 30 is invalid, infra, we need not resolve these addi- tional issues. Case: 24-1078 Document: 46 Page: 7 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 7

the inventors were in possession of that full subgenus, in- cluding the specific embodiment at issue in this litigation, bimatoprost. According to Sandoz, the specification does not provide a single example of an actual compound claimed by claim 30 and also fails to identify sufficient com- monalities of structure to provide a skilled artisan with the necessary “blaze marks” to lead them to the claimed com- pounds. Because we agree with Sandoz that no reasonable juror could have found that Sandoz failed to prove, by clear and convincing evidence, that claim 30 lacks adequate written description, we reverse the district court’s judg- ment. The written description requirement “reflects the basic premise of the patent system[:]” an inventor may “obtain[] a patent” only if she “discloses [the] invention” to the pub- lic, in sufficient enough detail that a person of ordinary skill in the art will understand that the inventor truly “pos- sessed the invention as claimed.” Regents of the Univ. of Minn. v. Gilead Scis., Inc., 61 F.4th 1350, 1355 (Fed. Cir. 2023). “The purpose of the written description requirement is to prevent an applicant from later asserting that he in- vented that which he did not.” Id. (quoting Amgen Inc. v. Hoechst Marion Roussel, Inc., 314 F.3d 1313, 1330 (Fed. Cir. 2003) (internal quotation marks omitted). Sandoz contends that the specification of the ’270 pa- tent is written so broadly that it encompasses a “universe of billions of compounds,” while claim 30 is limited to roughly 1,620 of these potential compounds. According to Sandoz, the patent fails to provide skilled artisans with the information they need to identify and obtain this subgenus of claimed compounds. Based on the vast discrepancy be- tween the written description and the actual scope of the claim, Sandoz continues, the pertinent artisan would con- clude that the named inventors did not actually possess the invention claimed in claim 30. Case: 24-1078 Document: 46 Page: 8 Filed: 11/18/2025

8 DUKE UNIVERSITY v. SANDOZ INC.

Certain premises underlying Sandoz’s written descrip- tion contentions are uncontested. Allergan’s expert, Dr. Reitz, agreed at trial that the specification is written in such a manner – with multiple open positions on its chem- ical structure, each of which could be filled in various ways – that it could encompass billions of compounds. See J.A. 10708 (Dr. Reitz agreeing “billions of compounds are represented by [the specification’s] backbone, based on the types of substituents”). Experts for both sides agreed that the number of compounds actually claimed by claim 30 is far smaller: either 4,230, which was Allergan’s expert’s view, or 1,620, as Sandoz’s expert opined. J.A. 10676, 10690-91, 10707. Thus, any reasonable juror had no choice but to find that the specification broadly described billions of compounds, while claim 30 was directed to between just 1,620 and 4,230 of such compounds. In order to have ade- quate written description, the specification of the ’270 pa- tent needs to allow a skilled artisan to understand how to identify this subgenus of claimed compounds. We have considered similar patents with some fre- quency. As we have previously held: Written description of an invention claimed as a ge- nus of chemical compounds, as here . . . requires description not only of the outer limits of the genus but also of either a representative number of mem- bers of the genus or structural features common to the members of the genus, in either case with enough precision that a relevant artisan can visu- alize or recognize the members of the genus. Regents, 61 F.4th at 1356; see also Ariad Pharms., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1350-52 (Fed. Cir. 2010). In other words, the specification must provide sufficient in- dication as to how a skilled artisan would narrow the dis- closed universe of billions of compounds described in the specification to the subset of just 1,620-4,230 compounds actually claimed. Case: 24-1078 Document: 46 Page: 9 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 9

Based on the evidence introduced at trial, no reasona- ble juror could find anything other than clear and convinc- ing evidence that the ’270 patent fails to describe either (i) a representative number of species of claim 30’s subgenus or (ii) structural features common to all members of that subgenus. Allergan does not even argue that the patent satisfies the first of these alternatives, since the ’270 patent does not expressly disclose even a single embodiment of claim 30. See, e.g., ’270 pat. at col. 8 (showing that only disclosed embodiment containing C(O)NHR3 at C1 end and phenyl in Z position does not contain methyl, ethyl, or iso- propyl in R3 position, thus falling outside of claim 30); see also J.A. 7939 (jury instruction, which was limited to com- mon structural features test); J.A. 10743-47 (parties dis- cussing whether jury should be instructed on representative number instruction, with court stating “there’s been no evidence that this . . . theory” applies.). Allergan solely relies on the proposition that the patent nonetheless sufficiently discloses structural features com- mon to all members of the claimed subgenus. To meet the written description requirement in this manner, the patent must “provide sufficient blaze marks” to direct a skilled ar- tisan to the claimed subgenus. Regents, 61 F.4th at 1356- 58. That is, the specification must “provide[] adequate di- rection which reasonably would lead persons skilled in the art to” the compounds actually claimed in claim 30. Fu- jikawa v. Wattanasin, 93 F.3d 1559, 1570 (Fed. Cir. 1996) (internal brackets and quotation marks omitted). Allergan insists that the specification of the ’270 patent satisfies this common-structural-features test because it discloses three features that are common to all members of the claimed subgenus: “[i] the characteristic prostaglandin hairpin, [ii] . . . with amides at the C1 position . . . [iii] con- nected to the unsubstituted phenyl ring at the omega [ac- tion] end.” Ans. Br. at 49 (bracketed numerals added). No reasonable juror could agree. Case: 24-1078 Document: 46 Page: 10 Filed: 11/18/2025

10 DUKE UNIVERSITY v. SANDOZ INC.

As we explain in more detail below, the ’270 patent, at best, discloses two prostaglandin hairpin structures and a menu of available atoms, moieties, and functional groups from which a skilled artisan could populate the R1, R2, R3, R4, X, Y, and Z positions of those structures. ’270 pat. at 7:34-8:67. Under our precedent, this is inadequate. See, e.g., In re Ruschig, 379 F.2d 990, 995 (CCPA 1967) (finding inadequate written description where specification leaves skilled artisan needing “to select[] from the myriads of pos- sibilities encompassed by the broad disclosure, with no guide indicating or directing that th[e] particular selec- tion[s] should be made rather than any of the many others which could also be made”). Start with the prostaglandin hairpin. It was undis- puted at trial that this is a “generic” feature. See J.A. 10670 (Allergan’s Dr. Reitz); J.A. 10574 (Sandoz’s Dr. Hla); J.A. 10486 (Sandoz’s Dr. Heathcock: “That’s the basic carbon skeleton of all prostaglandins.”). Both parties’ experts agreed that the specification discloses a character- istic prostaglandin hairpin structure shared by all com- pounds disclosed in the specification. See J.A. 10517-18 (Dr. Heathcock agreeing specification contains “a charac- teristic backbone . . . with an R group attached at C1”); J.A. 10670 (Dr. Reitz explaining prostaglandin has “char- acteristic hairpin turn, including the linker,” with R1 being “the important position”); J.A. 10668 (Dr. Reitz describing hairpin as “common”). Billions of compounds contain this generic hairpin structure. J.A. 10708 (Dr. Reitz testifying that “billions of compounds are represented by th[e] back- bone”). Allergan failed to identify how this common struc- tural feature was unique to the claimed subgenus, as opposed to the entire genus described in the specification. Thus, Allergan has not introduced evidence to show that a person of ordinary skill in the art would be able to “visual- ize” the thousands of compounds claimed in claim 30, from among the billions of prostaglandin compounds described Case: 24-1078 Document: 46 Page: 11 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 11

in the specification, based on the written description of the widely shared hairpin structure. Second, the hairpin structure does not provide suffi- cient guidance to account for the variation at the R1 and Z positions. A reasonable juror would necessarily have found that the specification fails to provide sufficient blaze marks with respect to the C1 position. Allergan argues that “the specification undisputedly discloses only 13 options for [this] action end,” with one of those options being the claimed amide, C(O)NHR3. Ans. Br. at 51 (emphasis in original). But the characterization of the specification’s disclosures for C1 as containing merely “13 options” is not correct, as most of the 13 options are themselves large cat- egories within which numerous options exist. In reality, the specification’s guidance with respect to the C1 position resembles a path with 13 branches, and most of those branches lead to additional branches, yielding in the end a vast number of options for C1. 3

3 Allergan insists that Sandoz’s expert, Dr. Heathcock,

“agreed that the specification discloses only 13 possibilities for the action end, one of which is an amide.” Ans. Br. at 17 (citing J.A. 10518-19) (internal quotation marks omit- ted); see also id. at 55 (asserting that “[Dr.] Heathcock re- peatedly conceded” this point). No reasonable juror could have found that Dr. Heathcock admitted the 13 options listed are only 13 and do not have embedded within them additional options. At no point, even in the cross-examina- tion highlighted by Allergan, did Dr. Heathcock testify there are “only” 13. To the contrary, he consistently made clear that while he could visualize each of the “functional groups” listed for R1 (i.e., C1), he further explained that it “would be hard, because there’s so much variety with the Rs.” J.A. 10519; see also J.A. 10520 (Allergan counsel ac- knowledging Dr. Heathcock “said there’s variability in these– in each one of these 13”). Case: 24-1078 Document: 46 Page: 12 Filed: 11/18/2025

12 DUKE UNIVERSITY v. SANDOZ INC.

In particular, the specification teaches: R1 is selected from the group consisting of [1] C(O)OH, [2] C(O)NHOH, [3] C(O)OR3, [4] CH2OH, [5] S(O)2R3, [6] C(O)NHR3, [7] C(O)NHS(O)2R4, [8] tetrazole, [9] a cationic salt moiety, [10] a pharma- ceutically acceptable amine or [11] ester compris- ing 2 to 13 carbon atoms, and a [12] biometabolizable amine or [13] ester comprising 2 to 13 carbon atoms. ’270 pat. at 8:9-15 (internal numbering added). A skilled artisan would understand that only four of these 13 options (C(O)OH, C(O)NHOH, CH2OH, and tetrazole) are singular items; the other nine options are categories, each requiring additional choices within them. Thus, selecting one of Al- lergan’s “13 options” would, for nine of the branches on the path, be merely the first decision an artisan would need to make before having to make other decisions just to deter- mine what to place at C1. For example, one of the patent’s disclosed options (number [3] above) is C(O)NHR3. The inclusion of the sub- stituent R3 as a component of this option requires a further choice to be made (after choosing C(O)NHR3 from among the 13 options for C1), because the specification discloses no fewer than 12 further categories from which the artisan must select to fill the R3 position. See ’270 pat. at 8:10-17; see also id. at 5:22-36 (defining one of the 12 categories, the “heterogenous group,” as “a saturated or unsaturated chain containing 1 to 18 member atoms,” which may be straight or branched in multiple spots, with or without (double and triple) bonds). Hence, even assuming a person of ordinary skill chose C(O)NHR3 from among the “13 options” for C1, just to fill in the C1 position she would still have to (i) select 1 of 12 categories for the R3 position, and then (ii) select a specific molecule from whichever category of those 12 she chose. While Dr. Reitz told the jury “any ordinary chemist could look at the functional groups, understand what’s Case: 24-1078 Document: 46 Page: 13 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 13

intended, understand permutations that could result,” J.A. 10670, and therefore might know not to pursue every one of the possible combinations the specification allows to be placed at C1 and its R3 component, the incontestable fact is that the specification describes far more than just “13 options” for C1. And as we have previously held: “Following [such a] maze-like path, each step providing multiple alternative paths, is not a written description of what might have been described if each of the optional steps had been set forth as the only option.” Regents, 61 F.4th at 1357. Or, as we have similarly put it, “one cannot disclose a forest in the original application, and then later pick a tree out of the forest and say ‘here is my invention.’” Purdue Pharma L.P. v. Fauld- ing Inc., 230 F.3d 1320, 1326 (Fed. Cir. 2000). Instead, sat- isfaction of the written description requirement with a patent like the ’270 patent requires disclosure of sufficient “blaze marks directing the skilled artisan to that tree.” Id.; see also Regents, 61 F.4th at 1356. The specification does not do so, and therefore fails to adequately direct the skilled artisan to the proper selection at the C1 position in order to arrive at the subgenus of compounds claimed by claim 30. Furthermore, the only blaze marks provided by the specification for this C1 selection point away from the com- binations as recited in claim 30. A specification’s descrip- tion of “preferred embodiments do[es] blaze a trail through the forest.” Fujikawa, 93 F.3d at 1571. The specification of the ’270 patent calls out five of its 13 options for C1 as either “prefer[red]” or “more prefer[red].” ’270 pat. 8:10-18. Notably, however, those five do not include C(O)NHR3, which is the option a skilled artisan would need to choose to reach the claimed invention. This means that the “pre- ferred” and “more preferred” blaze marks direct a skilled artisan away from, rather than toward, the claimed subge- nus, which would again lead the artisan to conclude the in- ventors did not actually possess what they claimed. See Case: 24-1078 Document: 46 Page: 14 Filed: 11/18/2025

14 DUKE UNIVERSITY v. SANDOZ INC.

generally J.A. 10454 (Dr. Heathcock opining that specifica- tion describes carboxylic acid and esters as more preferred and does not note any preference for C1-alkylamides like bimatoprost). If a skilled artisan were to ignore these blaze marks preferring five of the other “13 options” for C1, and remain on the nonpreferred C(O)NHR3 path, the specification then provides help by teaching that R3 in this group is “prefera- bly . . . selected from the group consisting of methyl, ethyl, and isopropyl,” ’270 pat. at 8:30-31. But Allergan has not identified any persuasive reason why an artisan who ig- nored the first set of blaze marks, effectively directing her away from C(O)NHR3, would then follow the blaze marks given for the R3 location of this non-preferred group. In a final effort to identify part of the written descrip- tion that would assist an ordinarily skilled artisan at the C1 position, Allergan points to portions of the specification that teach methods for compound synthesis, including four examples of synthesizing compounds containing amides. Allergan argues this discussion would steer a skilled arti- san to the placement of an amide at C1, and therefore di- rect that artisan to the right choice, C(O)NHR3, out of the alleged “13 options” elsewhere called out (which we ad- dressed above) as available for C1. See J.A. 10673-74 (Dr. Reitz testifying that specification points to amides at C1 because of their mentions in description of synthesis process); J.A. 10451 (Dr. Heathcock agreeing these disclo- sures are in specification); see also J.A. 10525 (Dr. Heath- cock: “They definitely say [in the patent] that you can make an amide, but they didn’t really give any explicit instruc- tions on how to do that, although they did for sulfona- mide.”). In each of the four synthesis schemes disclosed, however, C1 can be any of the same variants listed above – that is, any of the “13 options,” nine of which are categories requiring further embedded choices – but one ends up with an amide only if one selects option number six from that list, C(O)NHR3. And while the syntheses described as Case: 24-1078 Document: 46 Page: 15 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 15

Formulas I and III of the specification call out amides as examples of the optional C1 manipulation, even these for- mulas do not indicate that amides should be preferred at C1. Moreover, Allergan’s expert, Dr. Reitz, admitted that two specific examples provided in the synthesis schemes, which use sulphonamides or hydroxamic acid, are not within the scope of the ’270 patent’s claims. J.A. 10696-97. Finally, we turn to Allergan’s arguments about what the specification describes at the Z position. Claim 30’s compounds require a phenyl group at Z. J.A. 10675. While Allergan contends that “the specification discloses only eight categories of options for Z, and . . . expresses a pref- erence for phenyl,” Ans. Br. at 55, in fact the specification makes clear that (as with C1) each of these eight categories requires additional embedded choices: Z is selected from the group consisting of a carbo- cyclic group, a heterocyclic group, an aromatic group, a heteroaromatic group, a substituted car- bocyclic group, a substituted heterocyclic group, a substituted aromatic group, and a substituted het- eroaromatic group. ’270 pat. at 8:48-53 (emphasis added). While the specifica- tion does identify phenyl as the most preferred aromatic group, see id. at 4:2-3, this guidance is only pertinent once the artisan selects an aromatic group from among the eight initially described options, which nothing in the specifica- tion directs such an artisan to do. See J.A. 10675 (Dr. Reitz opining that when “Z is selected from an aromatic group,” a skilled artisan “would know that a phenyl is [an] aro- matic group” and “the most preferred aromatic group is Case: 24-1078 Document: 46 Page: 16 Filed: 11/18/2025

16 DUKE UNIVERSITY v. SANDOZ INC.

phenyl,” but not identifying a reason why an artisan would select an aromatic group in the first place). 4 Allergan’s argument that “at least ten of the patent’s example compounds employ unsubstituted phenyl at the omega end,” Z, Ans. Br. 52, does not overcome these prob- lems. The specification lists 95 example compounds and gives no reason to prefer the ten examples containing phe- nyl. Ten is not even the greatest number of appearances of a compound at the omega end; flurobenzene is used at Z in 18 of the 95 example compounds disclosed in the patent. Thus, even accepting that the specification guides a skilled artisan towards the hairpin structure, leaving only variability at the C1 and Z ends to be navigated, the speci- fication does not provide sufficient “blaze marks” to guide a skilled artisan to a PGF analog with an amide at the C1 position and a phenyl at the Z position, which are both re- quired elements of the compounds comprising claim 30’s subgenus. To the contrary, as in Fujikawa, 93 F.3d at 1571, the specification of the ’270 patent does not “direct one to the proposed tree in particular, and does not teach the point at which one should leave the trail to find it.” In- stead, the specification may only reasonably be viewed as a mere “‘laundry list’ disclosure of every possible moiety for

4 The specification further recites a preference for Z be-

ing selected from among a “group consisting of furanyl, thienyl, and phenyl.” ’270 patent at 4:59-63. However, for that blaze mark to be pertinent to the skilled artisan, she would need to have already chosen a —C=C— bond for the X position. See id. at 8:54-63. Such a —C=C— bond is just one of as many as 15 options for X, see id. at 8:41-47; and it is the only one that if selected contains a blaze mark point- ing to phenyl. Therefore, the skilled artisan must first reach a conclusion regarding X, the linker, before the spec- ification may, but in most cases will not, prompt her to pre- fer phenyl. Case: 24-1078 Document: 46 Page: 17 Filed: 11/18/2025

DUKE UNIVERSITY v. SANDOZ INC. 17

every possible position,” making it inadequate to satisfy the written description requirement of 35 U.S.C. § 112(a). See id. Thus, this is a case in which the appellant has over- come the doubly high burden of persuading us to overturn a jury verdict of no invalidity. The multiple, branching paths of the ’270 patent’s specification are clear on the face of the patent, were explained in detail by Sandoz’s expert, and their existence was not disputed by Allergan’s expert. Given our precedent, any reasonable juror would have found, by clear and convincing evidence, that a person of ordinary skill in the art, reviewing the specification of the ’270 patent, would be unable to visualize or recognize the members of the subgenus claimed by claim 30 based upon the specification’s disclosures. The specification fails to provide the relevant artisan with sufficient blaze marks or structural commonalities among the claimed compounds to lead her to conclude that the inventor actually possessed the claimed invention. Therefore, we reverse the judgment of the district court that claim 30 is not invalid for lack of a written description. See, e.g., BASF Plant Sci., LP v. Commonwealth Sci. & In- dus. Rsch. Org., 28 F.4th 1247, 1269 (Fed. Cir. 2022) (re- versing judgment of adequate written description where “jury had no reasonable basis to reject [accused infringer’s] evidence . . . of inadequate written-description support”); Idenix Pharms. LLC v. Gilead Scis. Inc., 941 F.3d 1149, 1164 (Fed. Cir. 2019) (reversing same where patent “fail[ed] to provide sufficient blaze marks to direct a POSA to the specific subset” of claimed compounds). IV We have considered Allergan’s remaining arguments with respect to written description and find them unper- suasive. Accordingly, for the reasons set out above, we con- clude that no reasonable factfinder, taking the evidence in the light most favorable to Allergan, could have found Case: 24-1078 Document: 46 Page: 18 Filed: 11/18/2025

18 DUKE UNIVERSITY v. SANDOZ INC.

anything other than clear and convincing evidence that claim 30 of the ’270 patent is invalid for lack of adequate written description. The judgment of the district court is reversed. REVERSED