Reading the Spec, Not the Press Release
The second in a five-part series of articles about how biopharma BD&L should use the patent record in a post-FOIA world.
A patent specification is the only document in a biopharma alliance whose author has been forced, under threat of invalidity, to disclose what the asset actually is. The press release is marketing. The deal announcement is negotiated optics. The S-1, when there is one, is curated narrative. The patent specification — the abstract, the claims, the description, the working examples, the sequence listings — is the one place where the inventor and the prosecuting attorney had to write down, in front of a patent examiner, what was new, what was enabled, and how broadly the new thing could be claimed without collapsing under the weight of its own ambition. Forty years of FOIA-released alliance contracts let BD&L professionals triangulate around that specification. Now the contracts are gone, but the specifications are still there. Reading them is no longer a supplementary exercise. It is the exercise.
This past April, The Economist ran a piece on Scribe Therapeutics and the broader push toward epigenome editing as the next translational wave after the first CRISPR knockouts. The article is well-written and the underlying science is real: editors that switch genes off rather than cutting DNA, with the prospect of durable disease modification without permanent genomic alteration. As topical biopharma writing goes, it is on the leading edge. As a basis for a partner-of-choice or asset-valuation decision, however, it is exactly what every BD&L professional has learned to be careful around — a press-release-shaped story about a single company with 43 published patents, told without reference to the 700-plus other published epigenome editing patents that the same prospective partner would also have to clear, license around, or defend against.
That broader universe is what a published patent record makes available, and what reading specifications rather than press releases is for.
The exercise I want to walk through here used 788 biopharma patents in epigenome editing, CRISPR-based gene regulation, and adjacent gene/cell therapy filings since 2022 — the published-patent universe that any partner doing serious diligence would have to read in full. The full-text specifications were scored against a structured rubric covering six dimensions every BD&L professional already evaluates qualitatively: how specifically the abstract names a target and a clinical positioning; whether the independent claims are anchored in structure or just in function; how the dependent claims fall back when the independent claim is challenged; what the prior-art neighborhood looks like; whether the working examples actually enable the claimed scope under Amgen v. Sanofi; and how broadly the claims could survive a real opposition or Inter Partes Review (IPR). The top decile — 79 patents — was then put through a complete pairwise tournament, every patent scored head-to-head against every other patent in the cohort, 3,081 pairs in all. The output is not a leaderboard of companies. It is a map of where the foundational claim density actually sits.
The map looks roughly like the field’s reputation, with three commercial filers worth pulling out.
The reputation part: at the top of the top decile sit the academic and academic-affiliated holders that anyone in the field would expect. The Liu lab at Harvard on phage-assisted continuous evolution of base editors. The Broad Institute and MIT on Cas variants, type II and type VI systems, prime editor architectures, and engineered AAV capsid delivery. Whitehead on programmable epigenetic modification. The Regents of UC and the Charpentier filings on the original CRISPR-Cas9 platform. The single highest-ranked patent in the cohort — a 196,000-word Liu-lab specification on PACE-evolved base editors — finished the pairwise tournament at 78-0. Not because it claims everything; because everything in the field has to be claimed around it. That is what foundational IP looks like when you read the spec rather than the press release.
The exceptions are where the BD&L decision actually gets made.
While not highlighted in the Economist piece, Beam Therapeutics has the deepest commercial top-decile presence in the analysis — six patents in the top 79, with calibrated ensemble means that hold across the cohort rather than collapsing to one or two strong filings and a long tail. That is the structural signature of a company that has been prosecuting on multiple fronts since founding rather than relying on a single platform claim and a press release. Beam’s pattern is what good commercial epigenome IP looks like at scale.
Scribe, by contrast — the company The Economist did highlight — has two top-decile patents on PCSK9-targeted CRISPR compositions that land at ranks 51 and 52 of 79. Inside the top decile of the 788-patent universe, but in the bottom third of the top decile, and the rubric’s text-grounded justification for both is the same phrase: “crowded prior-art subfield; novelty calibrated below pioneers.” Both patents are competently drafted and have the structural-disclosure depth one expects of a serious filer (113,000-word specifications, full sequence listings, working examples). Neither is a foundational claim. The press-release version of Scribe — that the company is at the leading edge of epigenome editing and PCSK9 represents a high-value translational application — and the spec version — that the company’s claims sit downstream of a dense academic prior-art ladder it does not own and will have to license around or design around — are both true. Only one of them is a reliable basis for setting an upfront, a royalty, or an asset price.
Another named company in the Economist piece sits in a similar place. Its single top-decile patent, on HBV gene repression, ranks at #66 of 79. The specification is substantial — 132,000 words, 453 claims, nearly 1,000 sequence IDs — and the rubric flags the same crowded prior-art subfield, with calibration haircuts for thin claim diversity and a shallow dependent-claim hierarchy beneath the broad genus the independent claim attempts. Volume of disclosure is not the same thing as foundational position, and the spec shows where the difference lies.
The throughline runs back to a problem the prior series spent considerable time on: the gap between what the press release says a deal is worth and what the contract says it is worth. The multiple is always there. Before 2020, FOIA-released contracts let BD&L professionals see how that multiple actually settled across hundreds of similar deals. After 2020, the multiple is still there, but the contract record that disciplined it is gone. What survived is the specification. The spec doesn’t tell you the upfront. It tells you whether the upfront the press release implies is supported by claims that would actually clear an opposition, an IPR, or a jury — or whether the upfront is supported by a press release.
A BD&L professional who reads the spec is, in the post-FOIA world, doing what an experienced BD&L professional has always done: looking past the announcement to the underlying instrument. The instrument is just different now. The press release will tell you what the company wants the deal to look like. The patent will tell you what the deal can survive being.
This is the second of five articles on what biopharma BD&L practitioners can read from the patent record now that the FOIA-disclosed contract record has ended. Here’s the start of the series .
The next article looks at how the patent record reshapes partner-of-choice analysis in a defended therapeutic area — using the GLP-1 / dual-agonist / triple-agonist incretin field, where the foundational filings span fifteen years and the current foundational filing layer is being written by a different cohort than the press-release version of the field describes.