Reading the Wake of a Surprise

There is an awkward question every BD&L professional eventually has to ask: How did I not see this coming? Not the compound/indication/marketing juggernaut rolling downhill, but the clinical outcome that wrong-footed the entire industry. For the GLP-1 / dual-agonist / triple-agonist class, the wrong-footing event was the 2017 indication shift — a belated recognition that the diabetes peptide franchise Novo Nordisk and Eli Lilly had quietly built through the late 2000s and early 2010s had become the foundation of what is now the largest obesity-metabolic disease market in the history of the biopharma industry.

Transactions that have commenced over the past eighteen months are still working through the consequences of that shift. On June 1 of this year, Lilly announced a $1.2 billion alliance with Hanmi Pharmaceutical centered on a GLP-2 program for short bowel syndrome; in October 2025, Novo announced a $5.2 billion acquisition of Akero Therapeutics for an FGF21 analog in metabolic dysfunction-associated steatohepatitis (MASH); in the same month, Pfizer announced a $10 billion acquisition of Metsera for a once-monthly GLP-1 / glucagon dual agonist. Three transactions; three different GLP-1-adjacent mechanisms; one shared theme, which is that the obesity-metabolic disease franchise has decisively moved off its diabetes-only footing and that adjacent-mechanism IP is being valued at multiples that would have looked impossible three years ago.

A reasonable back test of any putative “IP predictor” of changes in biopharma’s commercial landscape is to ask what it would have said about this GLP-1 class in late 2016, a year before the indication shift began to be visible in clinical readouts and four years before semaglutide’s obesity label. Specifically, if we kept the KC scoring framework and prompts constant, froze the GLP-1 published IP portfolio at December 31, 2016, and pretended the past nine years of commercial validation had not yet happened — what would rigorous KC scoring say about whose GLP-1 IP was foundational and best-situated to benefit from the obesity-metabolic shift?

Five-hundred-eighty-four published GLP-1-related biopharma patents remain as the pre-2017 cohort. Run against these 584 patents, the KC scoring framework returned a top decile of 59 GLP-1 patents. The top-decile IP composition is informative on two distinct questions, which are worth considering separately rather than collapsing into a single “did it work” verdict.

The first question is what the KC scoring framework correctly identified about the protein chemistry it was actually scoring. Novo holds sixteen of the 59 top-decile slots, with the foundational semaglutide-family publications US8815802B2, US20130053311A1, and US8648041B2 sitting at calibrated ranks #1, #2, and #5 — the tournament rank #1 with a 57-1 pairwise record across the 58 head-to-head comparisons, close to a clean sweep of the substrate’s top decile. The Indiana University filings from Richard DiMarchi’s laboratory contribute five top-decile filings on glucagon-superfamily peptides exhibiting nuclear hormone, glucocorticoid, and triple-agonist activity; that is a 3.15× enrichment over the cohort’s baseline rate. Zealand contributes three top-decile filings, including the gastrin-conjugate dual agonist US9089538B2 at calibrated rank #4. Amylin’s polypeptide-conjugate triple agonist US20140066370A1 sits at rank #6. Sanofi-Aventis Deutschland contributes the prodrug-insulin-conjugate composition US20120184489A1 at rank #19.

All told, six of the seven commercially-validated GLP-1 agonist programs have their composition-of-matter precursor work present in the top decile in some form — and across the dual-agonist and triple-agonist subbuckets, the platform IP that tirzepatide later validated commercially is enriched 3.15× over the cohort baseline. So, on the question of whether the KC scoring framework read claim craftsmanship, structural anchoring, dependent-fallback hierarchy, and prior-art neighborhood in a way that surfaced the foundational protein chemistry the field is now built on, the back-tested answer is “yes”.

The second question is what the framework did not predict, and that is the indication shift that resulted in today’s obesity-metabolic disease market. The dulaglutide precursor patents — Eli Lilly’s Glaesner-class fusion-protein filings from 2007 onward — landed at ensemble rank 193 (the 33rd percentile, in the substrate’s middle band). Lilly’s 2009 Fc-formulation continuation landed at rank 488. Lilly’s candidate third precursor landed at rank 509. None are top decile. The Hanmi LAPS-conjugate filings, sixteen of which appeared in the pre-2017 cohort, similarly didn’t enter the top decile; the highest LAPS filing reached ensemble rank 110. The pass/fail thresholds for these tests were pre-committed before the scoring run, so the dulaglutide and LAPS results are binary and not retrofitted. Why didn’t the Lilly and Hanmi pre-2017 patents score higher? The dulaglutide commercial success was driven by the convenience of weekly subcutaneous dosing in a crowded GLP-1 agonist field rather than by IP-quality dimensions. Similarly, the LAPS conjugate is a delivery technology whose value is in the formulation and the licensing arc. The KC scoring framework read these patents correctly as not-top-decile — what the framework measures is claim craftsmanship and structural rigor in the molecules being claimed, and so can’t be expected to predict commercial outcomes whose drivers are convenience, delivery, or indication-shift foreseeability.

That distinction is the one piece of the GLP-1 back-test result I want to emphasize, because it shapes how BD&L professionals should use this kind of IP structural input going forward. A patent-quality framework reads the IP that exists. It does not predict the future indications for which that IP will turn out to be useful. The Novo semaglutide pre-2017 IP that the KC scoring framework correctly identified as foundational was filed and prosecuted as type 2 diabetes IP; it became the foundation of the obesity-metabolic franchise because of clinical evidence that emerged years after prosecution choices were made and because of a regulatory pathway design that no patent-record analysis could anticipate.

The same will be true going forward. A rigorous IP scoring framework can tell the BD&L professional which filings in a 2026 target-disease landscape are the structurally rigorous foundational claims; it cannot predict which indication will, in 2034, turn those filings into a different franchise from the one currently contemplated. The 2017 GLP-1 indication shift was a clinical and regulatory event, not an IP event; the IP that mattered was already in place, but its eventual importance depended on data that had not yet been generated.

What the recent deal record does tell us — and what the back-tested KC scoring framework reading reinforced — is that the next indication-adjacent acquisitions in this class will be valued based on the structural rigor of the underlying IP, not against currently prevailing labels. Pfizer paid $10 billion for Metsera because the once-monthly dual-agonist composition is foundational on its own terms, irrespective of which metabolic indication carries the eventual peak revenue. Novo paid $5.2 billion for Akero because the FGF21 analog efruxifermin is foundational composition IP in a related metabolic mechanism, and Novo’s diligence presumably read the prosecution history with the same care it read its own semaglutide arc. The Lilly-Hanmi GLP-2 alliance is the inverse pattern: a delivery-platform partnership in a defined indication, priced as a delivery-platform partnership and not as a foundational composition acquisition. The patent record supports this read; the KC scoring framework, read against the pre-2017 cohort, would have supported it too.

The KC scoring methodology details — the six-component KC scoring rubric, the tournament-calibrated top-decile ranking, the construct-validity test for tirzepatide via the Indiana DiMarchi platform precursor, the test for dulaglutide and LAPS-conjugates — are written up in full at rDNA.ai, along with the back-tested top decile and the underlying KC scoring results. The point of running a back-test on GLP-1 was to subject the KC scoring framework to a result it could not retrofit. The cleanest signal is the one the scoring framework refused to give: it does not rank dulaglutide on IP-quality grounds alone, and that refusal is the test it had to pass to be trustworthy on the questions it does answer.

A patent landscape is, in the end, not a forecast. It is a map of work performed. The diligence work BD&L does in 2026 against a target-disease patent record will not predict a 2034 indication shift any more than the 2016 GLP-1 record predicted the post-2017 obesity-metabolic franchise. What a structural IP reading can do is keep the BD&L professional from confusing the convenience-driven commercial winner with the foundational-IP winner, and from paying foundational multiples for delivery-platform assets, or seeking delivery multiples for foundational ones. The wake of a surprise is readable; the next surprise, by definition, is not.

---

This is the sixth article in the rDNA.ai biopharma BD&L series.

Here’s the start of the series .