US 7695936

I'll search for authoritative information on this specific patent number.

Let me retrieve the actual claim text and check for any CAFC litigation.

Let me check for any 2026 appellate litigation and try to pull the granted claim set.

Summary — US Patent 7,695,936 B2

Bottom line: This is Genentech's foundational "knobs-into-holes" patent (the heterodimerization technology at the heart of essentially every Fc-based bispecific antibody). It is a continuation in the Carter/Presta/Ridgway family and issued in 2010 with a 1995 priority date. Its original term expired in December 2016, and I found no CAFC 2026 docket activity involving this number.

---

Bibliographic data (verified against the granted patent front page)

Sources: Google Patents; granted-patent PDF; patents-review.

---

Abstract (as printed on the patent)

"The invention relates to a method of preparing heteromultimeric polypeptides such as bispecific antibodies, bispecific immunoadhesins and antibody-immunoadhesin chimeras. The invention also relates to the heteromultimers prepared using the method. Generally, the method involves introducing a protuberance at the interface of a first polypeptide and a corresponding cavity in the interface of a second polypeptide, such that the protuberance can be positioned in the cavity so as to promote heteromultimer formation and hinder homomultimer formation. 'Protuberances' are constructed by replacing small amino acid side chains from the interface of the first polypeptide with larger side chains (e.g. tyrosine or tryptophan). Compensatory 'cavities' of identical or similar size to the protuberances are created in the interface of the second polypeptide by replacing large amino acid side chains with smaller ones (e.g. alanine or threonine). The protuberance and cavity can be made by synthetic means such as altering the nucleic acid encoding the polypeptides or by peptide synthesis."

(Note: the Google Patents OCR renders "heteromultimer" as "heteronutitimer"/"heteromultimeric" inconsistently; the correct term is heteromultimer.)

---

Independent claims — plain-language overview

⚠️ Important caveat: The granted patent has 73 claims. The sources I could retrieve give the claim text of the pre-grant publication US 2007/0014794 A1 (38 claims, same application), not a verified full listing of the 73 granted claims. The three independent claims below are quoted/paraphrased from that publication and are the substantive core that carried into the patent; I could not verify whether the granted set added further independent claims (e.g., a composition claim, a host-cell claim, or a multispecific-antibody claim) or whether the granted claim 1 was amended to add a hetero/homo-multimer ratio limitation (as appears in parent US 5,731,168). Treat the claim numbering below as the published-application numbering.

Claim 1 — method of making the heteromultimer (the core claim).
A method of preparing a heteromultimer made of a first and second polypeptide that meet at an interface, where the first polypeptide's interface has a protuberance ("knob") positionable in a cavity ("hole") in the second polypeptide's interface. The method has two steps:

• (a) culturing a host cell containing nucleic acid encoding both polypeptides, where the nucleic acid encoding the first polypeptide has been altered from its original sequence to encode the protuberance, or the nucleic acid encoding the second polypeptide has been altered to encode the cavity, or both, and the nucleic acid is expressed; and
• (b) recovering the heteromultimer from the host cell culture.

Claim 25 — the product (heteromultimer composition-of-matter).
A heteromultimer comprising a first and second polypeptide meeting at an interface, wherein the first polypeptide's interface comprises a protuberance positionable in a cavity in the second polypeptide's interface, and wherein the protuberance, the cavity, or both have been introduced (i.e., engineered/non-naturally occurring) into the respective interfaces.

Claim 35 — method of making, defined by side-chain volume changes.
A method of preparing a heteromultimer comprising: (a) altering a first nucleic acid so that an interface residue of the first polypeptide is replaced with a residue of larger side-chain volume, thereby generating a protuberance; (b) altering a second nucleic acid so that an interface residue of the second polypeptide is replaced with a residue of smaller side-chain volume, thereby generating a cavity, the protuberance being positionable in the cavity; (c) introducing both nucleic acids into a host cell and culturing so both are expressed; and (d) recovering the heteromultimer from the culture.

Dependent-claim themes (typical of the family): the interface is the CH3 domain of a human IgG1; protuberance residues imported from Arg/Phe/Tyr/Trp (most preferred Trp or Tyr); cavity residues imported from Ala/Ser/Thr/Val; the heteromultimer is a bispecific antibody, bispecific immunoadhesin, or antibody-immunoadhesin chimera; host cells and vectors; and pharmaceutical compositions.

Worked examples in the specification (useful for claim construction): the T366/Y407 reciprocal swap — T366Y in one CH3 and Y407T in the partner (Kabat numbering) — plus the second pair F405A / T394W, and T366W/Y407A; heterodimer yields reported at >80%, preferably >90%. Note this is the earlier mutation set; the now-industrially standard T366W knob with T366S:L368A:Y407V hole (and S354C/Y349C disulfide) belongs to the later WO 98/50431 / US 7,642,228 branch of the same family, not to the examples of the '936 patent — a distinction frequently blurred in later citing patents.

---

Discrepancies and uncertainties to flag

1. Priority date conflict. Google Patents and the patent's own cover page show 1995-03-01 (filing of 08/399,106). Third-party database Unified Patents lists the priority date for US-7695936-B2 as 1995-02-28. I did not resolve this; the cover-page/continuity data (Mar. 1, 1995) is the more authoritative.
2. Claim-set verification gap. As noted, my sources gave the published application's claims, not a confirmed listing of all 73 granted claims. The three independent claims above are reliable as to substance but I cannot guarantee their granted claim numbers or that they are exhaustive.
3. Title oddity. The granted title is "Knobs and holes heteromeric polypeptides," distinct from the sibling patents' title "Method for making heteromultimeric polypeptides."

---

Litigation / CAFC 2026 check

My searches of the USPTO-facing and litigation databases for "7,695,936" / "7695936" in a 2026 Federal Circuit or district-court context returned no results. I found no evidence of any CAFC 2026 docket entry involving this patent number. This is consistent with its status: it is marked "Expired – Fee Related" with an adjusted expiration of 2016-12-21, so it cannot be asserted for ongoing infringement after that date (only past damages within the limitations period would have been reachable, and any such appeal would be unlikely to be docketed in 2026). I could not perform a direct PACER/Docket search, so I state this as "no evidence found" rather than an affirmative verification that no such proceeding ever existed.

If you need the granted claim set verbatim or a PACER-confirmed litigation history, that requires a direct pull from USPTO Patent Center / the USPTO Patent Public Search full-text API and PACER — neither of which was retrievable through the searches available to me.

I'll search for litigation involving US patent 7695936.

Bottom line

I found no litigation (district court, ITC, or CAFC appeal) involving U.S. Patent No. 7,695,936 ("Knobs and holes heteromeric polypeptides," Genentech, Inc.; inventors Paul J. Carter, Leonard G. Presta, John B. Ridgway). I want to be explicit that this is a negative finding based on the searches I was able to run — not a certification that no case exists, because I could not get search results back from the paywalled docket sources (PACER/Docket Navigator/Unified's case-list portal) in this session.

What I searched and what came back

Because Unified's case-list page and PACER are behind dynamic/paywalled interfaces, a snippet-based search will not reliably surface every case in which a patent was asserted. My searches did not penetrate those.

Important caution on identifiers: several results involved different patents with visually similar numbers — notably 7,696,236 (Shilpa Medicare cases in D. Del., filed 2019-01-14, terminated 2022-04-19), 7,369,058 (Activision TV v. Mazda), and 7,679,058/7,765,414 (HTC/Apple). Per your instruction I did not treat any of these as litigation on 7,695,936. None of them is the patent at issue.

Why an absence of litigation is plausible here

• The patent is recorded as expired – fee related, adjusted expiration 2016-12-21 (Google Patents bibliographic data for US 7,695,936 B2). Any damages window has closed, which strongly reduces the incentive to assert it now.
• The '936 patent is a later member of Genentech's "knob-into-hole" CH3 heterodimerization family (including US 5,731,168, 5,807,706, and 5,821,333). Genentech has historically exploited this technology primarily through licensing (including non-exclusive licenses to bispecific-antibody developers) rather than assertion litigation. To my knowledge the family's enforcement history is thin, but I cannot confirm this with high confidence from the results obtained.

What I cannot confirm (stated explicitly rather than guessed)

I do not have high-confidence knowledge of any specific case caption, case number, jurisdiction, filing date, or disposition for '936, and I will not fabricate one. I am also not in a position from these results to rule out:

• an IPR/PGR or other PTAB proceeding on '936 that never produced a litigation hit;
• a district court case where '936 was one of many asserted patents and did not appear in the snippet-level sources I could reach;
• foreign (non-US) proceedings, which are outside the databases I queried.

Recommended verification steps

Since this is a clearance-style question, I would verify before relying on the negative result:

1. Unified Patents Portal — District Court case list (https://portal.unifiedpatents.com/litigation/caselist), search patent number 7695936 (this is the interface I could not reach; the search engine results did not expose its case list).
2. USPTO Patent Center / Patent Trial and Appeal Board (PTAB) API — check for IPR/PGR/CBM petitions against 7,695,936.
3. PACER / Docket Navigator / Lex Machina — full-text search on the patent number, then manually confirm the number is 7,695,936 and not 7,696,236, 7,369,058, or 7,679,058.
4. CAFC dockets — search for "7695936" to catch any appeal referencing the patent.

If you'd like, I can draft the exact Boolean queries for Lex Machina/Docket Navigator (e.g., patent_number = "7695936" AND court_type = "district"), or you can paste in the Unified/PACER result pages and I'll compile the full case table (plaintiff, defendant, jurisdiction, case number, filing date, status) from them.

I'll verify the structured block's default by searching for any AIA trial activity on this patent.

Let me check for PTAB activity tied to this patent and its family.

Let me check the sibling/continuation patents — a challenge to a family member would be the likeliest route to an '936 proceeding.

PTAB / AIA Trial Analysis — US 7,695,936 B2

Source of the proceedings list: the canonical structured block supplied from the USPTO Open Data Portal (ODP) ingest. Per that block, and per my own searches, the list is empty. Everything below is therefore a negative finding plus the strategic implications — not a set of proceeding write-ups.

---

Proceedings overview

There are zero AIA trial proceedings on file for US 7,695,936 — no IPRs, no PGRs, no CBMs, across all statuses (active / invalidated / sustained / settled / institution-denied), so there is nothing to break down by status. Defensive posture for a defendant: the patent was never stress-tested at the Board — but it also no longer needs to be, because it is bibliographically recorded as Expired – Fee Related with an adjusted expiration of 2016-12-21 (Google Patents bibliographic data). Any demand letter citing this patent today is asserting a dead property right; the practical answer is a term/expiration objection and a § 286 damages problem, not an IPR.

Why the empty list is credible here, not just an artifact

1. PGR was never available. Under 35 U.S.C. § 321, PGR requires an effective filing date on or after 2013-03-16. This patent claims a 1995-03-01 priority date (continuation chain to 08/399,106, now US 5,731,168 — see the "(63)" continuity data printed on the granted PDF). PGR is off the table by statute.
2. CBM was never a fit. CBM review under AIA § 18 was limited to claims directed to a "covered business method" of a financial product or service. A CH3-interface heterodimerization patent is outside that subject matter, and the CBM program sunset for new petitions in 2020.
3. IPR was the only available vehicle — and it was never used. IPR under § 311 covers any patent regardless of filing date, so an IPR petition on the '936 was always theoretically possible from 2012-09-16 onward. None was filed.

Search corroboration (and an important identity caveat). My searches for "7695936" IPR PTAB, "7,695,936" PTAB petition final written decision, and IPR Genentech "knobs into holes" surfaced no petition, institution decision, or FWD naming 7,695,936. What they did surface was extensive numerical confusion that a defendant-side practitioner must screen for: the "936 patent" in the AIA case law is US 7,110,936 (ComplementSoft, the SAS Institute v. Iancu partial-institution case), and there is a US 7,594,168 / US 7,6992xx-adjacent set of unrelated results. I confirmed none of these is US 7,695,936. Do not let a vendor's docket report conflate 7,695,936 with 7,110,936 — the similarity is coincidental.

For the record, the Genentech IPRs that are real and that a search will surface are on unrelated patents: IPR2017-01139 (US 6,627,196), IPR2017-01140 (US 7,371,379), IPR2016-01771 (US 7,622,115, bevacizumab), IPR2016-01837 (US 7,807,799, trastuzumab), and the trastuzumab-method patents US 7,846,441 / 7,892,549 litigated up to the Federal Circuit at Nos. 19-1263, 19-1265, 19-1267, 19-1270. None of those is in the knob-into-hole family, and none names the '936.

---

Strategic summary

Claim status: everything is UNTESTED, and nothing is CANCELED. No claim of US 7,695,936 has ever been adjudicated by the PTAB. There is no FWD to parse, no canceled independent claim, no surviving-after-IPR claim set to enumerate. The 73-claim granted set (per the front page) stands exactly as issued — but it stands on an expired patent. That distinction matters enormously: an untested claim set would ordinarily be a defendant's nightmare, but an untested claim set on a lapsed patent is largely an academic curiosity. The claims also were never narrowed by prosecution disclaimer in an IPR, so nothing about their scope has been judicially or administratively settled.

Estoppel landscape: there is none, in either direction. Because no IPR was ever instituted, § 315(e)(2) estoppel binds nobody. No petitioner, real party in interest, or privy is barred from raising any § 102/§ 103 ground in district court. Conversely — and this is the point most often missed — the patent owner gains no benefit from the Board either: there is no IPR FWD affirming patentability that could be used as a shield against a § 282 invalidity defense. If you are ever in a district court case on this patent (e.g., a stale declaratory-judgment posture or a past-damages claim), all prior art remains fully available, subject only to the normal § 282/§ 325(d) and IPR-timing considerations. There is also no § 315(b) one-year clock running against anyone, because that clock is triggered only by service of a complaint alleging infringement.

Pattern signals: no adversary has ever engaged this patent at the Board. No repeat petitioner (the hallmark of a serial-challenged patent). No defensive aggregator in the chain for this number — Unified Patents has no case list for US-7695936-B2. No PTAB appeal history, because there is no PTAB decision. The absence of IPRs is a genuine signal: well-asserted, commercially-blocking patents attract IPRs, and this one never drew a single petition in the ~4.3 years between the IPR regime's start (2012-09-16) and the patent's expiration (2016-12-21). The most likely explanation is that the technology was monetized through licensing rather than assertion (consistent with the family's licensing history), and that the commercially operative claims live in the siblings — US 5,731,168 / 5,807,706 / 5,821,333 and US 7,642,228 / 8,216,805 / 8,679,785 — rather than in this terminal continuation. Caveat I will state rather than paper over: I could not complete a family-wide PTAB check on those siblings within this session, so I cannot tell you whether 5,731,168 or 7,642,228 were ever petitioned. That is a distinct question from the one you asked, and you should have it verified separately if the family is material.

---

Recommended next steps

If you are a defendant facing a demand or complaint citing 7,695,936:

1. Lead with expiration, not invalidity. The patent is recorded as expired 2016-12-21 (Google Patents); the 661-day PTA is printed on the granted PDF front page. There can be no post-2016-12-21 infringement. Combine that with the § 286 six-year damages lookback measured from the complaint date: a complaint filed in 2026 could reach back only to roughly 2020-09 — entirely after the patent lapsed. On that reasoning the recoverable past-damages window appears to be closed, and the § 286 defense, not an IPR, is your case-dispositive motion. (Legal conclusion offered as analysis from the bibliographic record; verify the term calculation and any terminal disclaimer yourself before relying on it.)
2. Do not file an IPR. There is no live exclusionary right to invalidate, and an IPR on an expired patent gets Phillips claim construction with no amendment rights and no practical upside. Filing one would only generate § 315(e)(2) estoppel against you for no benefit.
3. Link the null result explicitly. The absence of any FWD means you cannot cite a claim-cancellation disposition — so say that affirmatively in your papers rather than implying one exists: "No AIA trial proceeding has ever been instituted against US 7,695,936." Ground that statement by pointing to the USPTO's own records at PTAB E2E and the PTAB API/decisions page at https://developer.uspto.gov/api-catalog/ptab-api, and to a null CourtListener query such as https://www.courtlistener.com/?q=%227695936%22.
4. Sibling check. Because priority, specification, and claim scope overlap, confirm the PTAB status of US 5,731,168, 5,807,706, 5,821,333, 7,642,228, 8,216,805, and 8,679,785 independently. Those share the same 1995-03-01 priority and therefore the same 2015/2016 term expiry, so the same expiration argument should travel — but I have not verified it for each.

If you are the patent owner (or its licensee) contemplating enforcement: there is no PTAB hurdle to clear, but there is also effectively no term left. The "knobs-into-holes" commercial position, if any remains, sits in the licensing of the later WO 98/50431-era claims (T366W knob / T366S:L368A:Y407V hole with the S354C/Y349C disulfide) in US 7,642,228 and its successors — as noted in the earlier sections of this analysis, that mutation set is not the mutation set exemplified in the '936 specification, which uses T366Y/Y407T and F405A/T394W. Do not blur those two in any enforcement or licensing representation.

I'll research the assignment record for US 7,695,936 through the available sources.

Assignment & Ownership Analysis — US 7,695,936 B2

Sourcing note up front (important, and it governs everything below): The USPTO Assignment Center (https://assignmentcenter.uspto.gov/, mirrored at https://assignment.uspto.gov/patent/index.html) is a JavaScript application that cannot be queried by the search tools available in this session. I ran multiple targeted searches against indexed copies of the assignment record (uspto.report, PEDS/ODP mirrors, Google Patents legal events) and no reel/frame entry for 7,695,936 surfaced in any of them. Google Patents' legal-events block for this patent (included in the authoritative text retrieved for this analysis) contains only prosecution/continuity events — priority claims to 11/533,709, publication, grant, the 12/700,618 and 13/494,870 child filings, and the 2016-12-21 adjusted expiration. It contains no assignment, merger, or change-of-name events.

I will not invent reel/frame numbers. Everything below is either (a) drawn from the granted front page and the patent text, (b) drawn from the Google Patents legal-events block, or (c) explicitly labeled as unverified.

---

Inventors

Pattern assessment — no adverse signal. All three are in-house Genentech scientists; the record shows no split inventorship, no non-Genentech co-inventor, and no evidence of an inventor retaining personal ownership. The one item worth noting neutrally: Ridgway was the junior/most transient of the three (a postdoc-era contributor who did not build a decades-long Genentech career, unlike Carter and Presta). A postdoc cycling out of a research institution is not a "fire-sale precursor" — it is the normal academic-to-industry transition. There is no evidence of all inventors departing within 12 months of filing; Carter and Presta in particular remained Genentech lifers (Carter later served as Genentech's VP of Antibody Engineering). Per your instruction not to infer from narrative, I flag this only to say the signal is absent.

⚠️ Do not conflate the 1995 inventorship with the 2006 filing. The three inventors are named on the original 1995 parent application (08/399,106), and the 2006 continuation 11/533,709 simply carries that inventorship forward. Any "departure" analysis should be run against 1995–1996, not 2006.

---

Original assignee

Genentech, Inc. (1 DNA Way / 460 Point San Bruno Blvd., South San Francisco, CA), a Delaware corporation.

• Primary line of business: biopharmaceutical R&D and manufacturing — a fully integrated, commercial-stage drug developer (Avastin, Herceptin, Rituxan, etc.).
• Current status: Operating. Genentech, Inc. continues to exist and operate as an independently managed member of the Roche Group; Roche has held a majority stake since 1990 and acquired the remaining minority interest in 2009. Genentech has never filed for bankruptcy and has never been party to a patent fire-sale. Note the corporate-perimeter point: the assigning/documenting corporate entity name has remained "Genentech, Inc." throughout, so there is no recorded Change of Name leg expected in this chain.
• Did Genentech ship products embodying the claims? Yes — strongly. This patent's disclosed subject matter is the "knobs-into-holes" CH3 heterodimerization platform, and Genentech/Roche commercialized it in Fc-based bispecifics (e.g., the anti-CD20×CD3 and anti-CD3×CD20 bispecific T-cell engagers in the Roche/Genentech clinical-commercial line). Genentech also monetized the family through non-exclusive out-licensing to third-party bispecific developers rather than assertion. This matters for the verdict: the owner is a practicing, revenue-generating operating company, which is the inverse of the NPE fact pattern.

---

Assignment timeline

No recorded post-issuance assignment for US 7,695,936 was located in any indexed source. I state this as a negative finding on the sources I could reach, not as a certification that the Assignment Center contains no records.

For completeness, the following is the only conveyance that must exist as a matter of record and law for this chain, presented with an explicit unverified flag because I could not retrieve reel/frame data:

• [execution date unverified, expected 1995] / recorded [unverified] — Reel ____/ (NOT RETRIEVED — do not cite)

  • Conveyance: Assignment of assignors' interest (expected)
  • Assignor: Paul J. Carter; Leonard G. Presta; John B. Ridgway (individually)
  • Assignee: Genentech, Inc.
  • Correspondent: not retrieved. ⚠️ I cannot report an attorney of record for the recording. Genentech's patent filings of this era conventionally used in-house Genentech Patent Legal / outside firms, but naming one without the reel image would be fabrication. Flag: unknown — must be pulled from the reel image.
  • Context: Employment/obligation-to-assign — the standard first leg of title in which employee-inventors assign to their employer. Not a fire-sale, not a transfer-to-asserter.

• No second leg found. No assignment, security agreement, merger, change-of-name, license recordation, or release appearing after grant was surfaced. The Google Patents legal-events block for this patent is silent on assignments.

Consequence of the negative finding: if, on verification, the only entry is the inventor→Genentech leg above, then Genentech, Inc. remains the owner of record to this day and no third party ever held title. This is consistent with — and corroborated by — an item already established in the prior sections: the patent is recorded as "Expired – Fee Related," adjusted expiration 2016-12-21, i.e. the maintenance fees simply stopped being paid. Only the beneficial owner pays maintenance fees. Fee-lapse is affirmative evidence that no assigning entity took over and had a commercial reason to keep the asset alive.

Corroborating public-company record (SEC)

Genentech's 10-K filings through the relevant period (and Roche's subsequent Form 20-F disclosures) are the place to confirm (i) that no Genentech patent portfolio — and specifically not the CH3 heterodimerization family — was sold, spun out, or contributed to a monetization vehicle, and (ii) whether any non-exclusive license revenue was recognized on this family. My searches did not surface any SEC disclosure of a divestiture of the '936 family, and Genentech's licensing of knobs-into-holes is historically characterized as non-exclusive out-licensing while retaining ownership — the opposite of privateering. (Stated with moderate confidence; I could not run a full-text EDGAR search in this session.)

---

Timeline diagram

timeline
    title Ownership of US 7695936
    1995 : Three Genentech inventors file parent application
    2006 : Continuation filed by Genentech
    2010 : Patent issues with Genentech as assignee of record
    2016 : Patent term ends after fee lapse

Parser caution observed: no commas, slashes, or parentheses in event text.

---

NPE / troll-pattern signals

---

Verdict

Insufficient data — with the strong qualifier that every affirmative indicator points away from an NPE.

This is the least-satisfying verdict label for what is substantively a clean record, so let me be precise about why it is nonetheless the correct call and what would change it. The choice is driven by an evidentiary gap, not by ambiguity in the facts: I could not open the USPTO Assignment Center in this session, and no reel/frame entry for 7,695,936 appeared in any indexed source I could reach, so I have zero confirmed conveyances — not even the inventor→Genentech leg with a citable reel/frame. Under your rubric, "no records, or only the original assignment" maps to Insufficient data, and I am honoring the constraint against fabricating assignments rather than inventing a reel/frame to fill the format.

What is solidly established, and is the substance of the finding: the only assignee ever named in connection with this patent — on the grant front page, in the Google Patents bibliographic record, and in the legal-events block — is Genentech, Inc., an operating pharmaceutical company that ships products practicing the knob-into-hole claims. The patent lapsed for non-payment of maintenance fees with a 2016-12-21 adjusted expiration, which is affirmative evidence that no monetization entity took title and had a commercial incentive to sustain the asset. All eight NPE signals are not present or, in the single case of correspondent data (signal 3), not assessable. On the record as it stands, this is a single-link, original-owner chain with no troll pattern.

Verification path (I recommend running this before relying on the negative result):

1. USPTO Assignment Center Quick Look-Up by patent number 7695936 and by application number 11533709 — https://assignmentcenter.uspto.gov/ (indexed mirror: https://assignment.uspto.gov/patent/index.html).
2. USPTO ODP / File Wrapper API, which returns assignment history per application: GET https://api.uspto.gov/api/v1/patent/applications/11533709/assignment (API key required; this is the machine-readable equivalent and would settle the reel/frame and correspondent questions definitively).
3. If the Assignment Center returns only the inventor→Genentech leg, the verdict should be reclassified to "Operating-company, non-asserting" — Genentech, Inc. as owner of record, no assertion activity — which is the accurate characterization and is fully consistent with everything above.

Discrepancy flagged (carry-over, affects the assignment-window framing only)

The prior sections already noted that Unified Patents lists the priority date for US-7695936-B2 as 1995-02-28, while the granted cover page and the continuity chain show 1995-03-01 (filing of 08/399,106). This does not change any assignment conclusion — it only shifts the earliest possible execution date for the inventor→Genentech leg by one day — but any date-window query against the Assignment Center should be run with both dates in mind.

I'll search for the specific patent and its cited prior art references.

Let me pull the precise cited-reference details from the patent's front page and verify the foreign references.

Prior Art Analysis — US 7,695,936 B2 ("Knobs and holes heteromeric polypeptides")

Scope and sourcing note

You asked me to search the USPTO database for 7,695,936 specifically and analyze its patent citations. What I was able to retrieve is the "References Cited" (56) section as reproduced on the granted patent's own front page (the patentimages PDF of US 7,695,936, which reproduces the USPTO-issued front page, surfaced via Google Patents). I could not directly reach USPTO Patent Center / Patent Public Search full-text in this session, so everything below is grounded in that front page plus the granted-patent text you supplied. I checked that every number below is exactly 7,695,936 — not 7,696,236, 7,369,058, or 7,679,058 (similar-looking numbers that appear in unrelated matters).

Important structural finding: the '936 front page cites a very thin patent-art list — one U.S. patent and two PCT publications. The substantive prior art the examiners relied on is almost entirely non-patent literature (NPL), ~45 references. I treat both, but the "patent citations" proper are only the three documents in the table immediately below.

---

1. Patent documents cited in US 7,695,936 (as printed)

⚠️ Identifier flag (per your "interpret literally / don't auto-correct" rule): the front page of the '936 renders the second PCT as WO 93/11162, but the closely-related sibling patent US 8,679,785 prints the analogous slot as WO 93/11167 (which is titled "Polymerisation process" and is topically unrelated to bispecific antibodies). I have not resolved this; I report the '936's rendering (93/11162), which is the topically coherent match, but the number should be confirmed against the printed patent.

Not cited in the '936 (do not attribute to it). The patents that show up in the sibling '785 listing — US 5,731,168, 5,807,706, 5,821,333, 7,183,076, 7,642,228, 7,951,917, 8,216,805, and U.S. pubs 2007/0098712 and 2007/0178552 — belong to the '785's "Referenced Cited," not the '936's. Most are Genentech family members, not prior art.

---

2. The §102 framework that governs these references

The '936 is a pre-AIA continuation claiming priority to March 1, 1995 (app. 08/399,106; the same family is sometimes databased as 1995-02-28 — a one-day discrepancy I flagged earlier and cannot resolve; it does not change the analysis). Two consequences:

1. Family members are not prior art. US 5,821,333 (item 1) shares the inventors, the assignee, and the 1995 priority with the '936. It is not "prior art" under §102(a), (b), or (e); it was cited for continuity/double-patenting context. Even if it were art, it describes the same invention, so it would anticipate everything — which is precisely why it isn't art.
2. Anything published after March 1, 1995 is not §102(b) art, and (absent a priority failure) not §102(a) art either. This is critical here, because a large share of the cited NPL post-dates the priority date (see §4 below).

---

3. §102 anticipation assessment for each cited patent document

US 5,821,333 — No anticipation. Not art (same inventive entity/family/priority). Cited for continuity only.

WO 92/10209 — §102(b) art (published Jun. 25, 1992, >1 yr before 3/1/1995), but no anticipation of the core claims.

• It discloses two Fab fragments each bearing a "complementary domain not naturally present in the Fab … capable of stably binding to the other but not to itself," forming a leucine zipper (claims 1, 4, 11). It does disclose the general concept of engineering a heavy-chain constant-region interface to drive hetero-association, and culturing a host cell (claims 5–8).
• It does not disclose the '936's defining feature: an interface residue replaced by one of larger side-chain volume to create a protuberance that is positionable in a compensatory cavity created by a smaller side-chain replacement at the partner CH3 interface. No amino-acid identity/position (e.g., T366/Y407) is disclosed.
• Therefore it cannot anticipate independent claims directed to the protuberance-into-cavity heteromultimer or method (published-app claims 1, 25, 35, and their granted counterparts). At most it anticipates a bare claim to "a bispecific antibody comprising two different Fab fragments associated through engineered heterodimerization domains" in a hypothetical claim broad enough to read on zipper-mediated association — but no such claim appears in the '936 family.

WO 93/11162 — §102(b) art (published Jun. 10, 1993), but no anticipation of the core claims.

• Discloses bispecific antibodies in which a first and second leucine zipper (Fos/Jun) join two epitope-binding components into an F(ab'-zipper)₂ heterodimer (claims 1, 11–16), including in vivo expression in a single cell (claim 16).
• Same gap as WO 92/10209: leucine-zipper complementarity is not a size-based protuberance/cavity engineered into an immunoglobulin CH3 interface by larger/smaller side-chain substitution. No anticipation of claims 1/25/35.

Net result: None of the three cited patent documents anticipates the '936's independent claims. They are §102(b)/§103 background art establishing that heterodimerization of antibody-like molecules was known; the '936's novelty lies in the size-complementary protuberance-into-cavity mechanism at the CH3 interface, which these references do not teach.

---

4. The references that actually carry the §102 weight are Non-Patent Literature

Because the patent-art list is so short, the examiner's substantive citations are NPL (all listed on the front page as "OTHER PUBLICATIONS," with a "cited by examiner" marker). The most §102-relevant ones:

Key §102 takeaway: The pre-1995 NPL (Crick, Milstein & Cuello, Brennan, Ward, Kostelny, O'Shea, Hollenbaugh, John, Shalaby, Chamow, Holliger, Fanger) is §102(b) art and collectively establishes that (i) bispecific/multispecific antibody assembly was known and (ii) engineering protein–protein interfaces (including complementary-residue and leucine-zipper strategies) was known. But none of them discloses a size-based protuberance-into-cavity at an IgG CH3 interface, so none anticipates the '936's independent claims. The references dated after March 1, 1995 (Belshaw 1995, McPhee 1996, Ridgway 1996, Zhu 1997, Thies 1999, Jefferis 1998) are not §102 prior art to claims entitled to the 1995 priority date — a point worth stressing for a continuation patent.

---

5. Honest limitations

• Granted claim set not verified verbatim. The '936 has 73 claims; I could not pull a confirmed full listing. My claim mapping uses the independent claims of the published application US 2007/0014794 A1 (claims 1, 25, 35 as the method/product method-by-side-chain-volume claims) as a proxy, plus the sibling '785's claim set. If the granted set added or narrowed independent claims, the anticipation conclusions for those specific claims should be re-run.
• Direct USPTO access not achieved. The "References Cited" list above is taken from the granted-patent front page (patentimages PDF), which reproduces the USPTO front page; a direct Patent Center / Patent Public Search pull to confirm the list letter-for-letter was not possible in this session.
• Two unresolved identifier/date discrepancies carried forward: the priority date (3/1/1995 on the cover vs. 2/28/1995 in some databases) and the WO 93/11162-vs-93/11167 rendering.
• No affirmative clearance. This is a reference-by-reference §102 mapping, not a freedom-to-operate or validity opinion.

Bottom line: For US 7,695,936, the patent citations are just US 5,821,333 (family member, not art), WO 92/10209, and WO 93/11162 — all leucine-zipper/complementary-domain bispecific-antibody art that is §102(b)-dated but does not anticipate the '936's protuberance-into-cavity claims. The examiner's substantive §102(b) art is the ~20 pre-1995 NPL references, none of which discloses the CH3 size-complementary knob/hole feature; the post-1995 NPL (including the inventors' own Ridgway 1996 paper) is not prior art to the 1995 priority date.

I'll pull the prior-art landscape for this patent before analyzing. Let me run targeted searches on the cited references and the family's prosecution history.

Let me pull the face-of-patent citations and the "Prior Art"/similar-documents data specific to the '936, plus the sibling Arathoon art.

Obviousness Analysis — U.S. Patent No. 7,695,936 B2 ("Knobs and holes heteromeric polypeptides")

0. Framing, and an important caveat about my source set

Provenance of the prior-art list I am using. The task directs me to "the results from the Prior Art section of this page." I must flag honestly: the Google Patents text supplied to me was truncated at the "Landscapes" section — the "Citations," "Cited By," "Similar Documents," and "Prior Art" blocks of the US7695936 page were not included in the material I received, and my follow-up search for that block verbatim returned only a one-line citation hit. What I was able to retrieve, and what this analysis is therefore built on, is:

• the face-of-patent citation/IDS lists for this family, as reproduced on FreePatentsOnline's page for sibling US 8,679,785 and for the parent US 5,731,168;
• the Google Patents bibliographic scrub showing the "Prior art date: 1995-03-01" and the assigned prior-art keywords ("polypeptide, heteromultimer, amino acid, residue, protuberance");
• third-party and citing-patent characterizations of the family (Unified Patents; several WO/JP family citations).

So the reference set below is a reconstruction from the family's own IDS/face citations, not a verbatim pull of the '936's own "Prior Art" section. Treat the identifications as reliable but the enumeration as possibly incomplete.

Claim text caveat (carried forward from the earlier Patent Summary). As flagged previously, I have reliable text for the published-application claim set (US 2007/0014794 A1), not a verified verbatim listing of the 73 granted claims. The independent claims analyzed below are the published-application independents (1, 25, 35), which are the substantive core.

---

1. The governing § 103 regime and the critical dates

The single most important structural point for this analysis: because the claims carry a 1995-03-01 effective date, the two publications most commonly associated with "knobs-into-holes" — Ridgway, Presta & Carter, Protein Eng. 9:617–621 (1996) (PubMed 8844834) and Atwell/Zhu/Carter (1997) — are not prior art. They are the applicants' own post-priority work. The same is true of Merchant et al., Nat. Biotechnol. 16:677 (1998) and Zhu et al., Protein Sci. 6:781 (1997). These appear in the family's IDS (and thus in the "Prior Art" display), but an IDS entry is not an admission that the reference is prior art, and a post-priority reference cannot be used under § 103(a). A challenger who builds a § 103 case on Ridgway 1996 has built it on sand.

Two further date traps worth naming because they appear in the family's citation lists:

• Belshaw et al., Angew. Chem. Int. Ed. Engl. 34(19):2129 (1995) — "Rational Design of Orthogonal Receptor-Ligand Combinations." Volume 34, issue 19 dates this to late 1995, i.e., almost certainly after the 1995-03-01 priority date. It cannot be § 102(b) art and is only § 102(a) art if it predates the actual invention — which is doubtful.
• McPhee et al., PNAS 93:11477 (1996) — cited in the parent's IDS, but post-priority. Not art.

The references that are available are listed in § 3.

---

2. The person having ordinary skill in the art (POSITA)

A POSITA here is a protein engineer/immunochemist with a Ph.D. and roughly 3–5 years of post-doctoral experience in antibody engineering and structure-guided mutagenesis, who:

• routinely reads Fc crystal structures and knows the immunoglobulin domain architecture (Deisenhofer 1981 was, by 1995, canonical);
• is competent in site-directed mutagenesis, cassette mutagenesis and mammalian transient/stable expression of antibody chains;
• is aware of the bispecific-antibody literature and its central practical failure — poor yield of the correct heavy-chain pairing.

This matters because almost every KSR rationale below turns on what such a person would consider a finite, identified, predictable design space.

---

3. The prior-art set available against the 1995-03-01 priority

(All items below are § 102(b) art unless marked; the sources are the family's own citation lists linked in § 0.)

(a) The problem: bispecific antibody production and random heavy-chain assortment

(b) The substrate: the CH3–CH3 interface and its structure

(c) The design principle: knobs-into-holes and interface remodeling for heterospecificity

---

4. Ground 1 — The core method claim (generic protuberance/cavity)

Claim at issue: published-app claim 1 (and its product/alternative-method siblings 25 and 31): a method of preparing a heteromultimer by culturing a host cell whose nucleic acid has been altered to encode a protuberance at the interface of the first polypeptide and/or a cavity at the interface of the second polypeptide, "or both," followed by recovery.

Proposed combination:
Deisenhofer 1981 + Miller 1990 (identity and structure of the CH3 interface) + Crick 1952/53 (knob-into-hole packing principle) + Ward 1987 (engineering complementary residues at a symmetric subunit interface to obtain heterodimerization) + Chamow 1994 (the co-expression/Ab-Ia-chimera production platform and the demonstration that wild-type CH3 gives a random mixture of Ab/Ia, IgG, and Ia homodimer).

Mapping:

Motivation to combine — and why it is strong on these facts:

1. The problem was explicitly identified in the art. Milstein & Cuello 1983 taught that co-expressing two different heavy chains gives a random mixture; Chamow 1994 measured that mixture directly in the very Ab/Ia system the '936 uses (the patent's own Example reports ~57% hybrid for wild-type CH3 — essentially the Chamow result). The Federal Circuit and the Board treat an art-recognized problem as itself supplying the motivation (KSR; MPEP 2144.04).
2. The art identified the precise molecular target. Deisenhofer and Miller told the POSITA that heavy-chain pairing is mediated by a well-defined, 16-residue-per-domain CH3 interface. This is the KSR "known structure ready for improvement" situation.
3. The art supplied the specific design principle. Crick's knob-into-hole packing is the named concept for complementary side-chain packing, and O'Shea 1991 had confirmed it crystallographically. Applying it to a symmetric homodimeric interface to force heterodimerization is precisely the "use of a known technique to improve a similar device in the same way" rationale (MPEP 2144.04(III)).
4. The art had already done the analogous thing. Ward 1987 (tyrosyl-tRNA synthetase), John 1994 (GCN4 zipper) and O'Shea 1993 (Peptide 'Velcro') collectively teach that complementary engineering of a symmetric interface converts homospecific association into heterospecific association. The claims do not require charge complementarity specifically; they claim any protuberance-into-cavity geometry, which is a broader concept than what Ward/John/O'Shea already achieved.
5. Expectation of success was reasonable. The Fc homodimer is symmetric and structurally permissive at the two central β-strands; the interface residues are aliphatic/aromatic and largely tolerant of conservative volume changes; and every element (mutagenesis, co-expression, protein A purification, SDS-PAGE quantitation) was routine.

KSR rationales available to an examiner: (A) combination of known elements with predictable results; (C) known technique improving a similar device in the same way; (D) applying a known technique to a known device ready for improvement; (F) "obvious to try" — a finite number of identified interface residues and a finite set of size-changing substitutions; (G) design incentive (the entire field was looking for it).

---

5. Ground 2 — Claims defined by side-chain volume, and the R/F/Y/W and A/S/T/V species

Claims at issue: published-app claim 35 (alter a first-interface residue to one of larger side-chain volume; alter a second-interface residue to one of smaller side-chain volume) and the dependent species claims reciting R/F/Y/W for the knob and A/S/T/V for the hole.

Proposed combination: Miller 1990 / Deisenhofer 1981 (which residues; their buried/contact status) + Ward 1987 + John 1994 + O'Shea 1993 (complementary interface engineering) + the standard side-chain-volume tables the '936 itself reproduces (Handbook of Chemistry and Physics, 43rd ed. (1961); Zamyatnin, Prog. Biophys. Mol. Biol. 24:107 (1972); Chothia, J. Mol. Biol. 105:1 (1975)) + Eriksson 1992 (cavity creation by large→small mutation).

Why obvious: the selection of R, F, Y, W as the largest side chains and A, S, T, V as the smallest is not a discovery — it is a reading off a published volume table. This is the paradigm of KSR's "finite number of identified, predictable solutions" and of MPEP 2144.04(VI). The patent's own specification concedes the point by expressly importing the Chothia/Zamyatnin volume tables rather than generating them.

A secondary, and legally potent, point: if the prior art combination inherently produces a greater heteromultimer:homomultimer ratio, then the "ratio … greater than that for a multimer which lacks the protuberance and/or cavity" functional limitation (present in the '168 claims and mirrored in the family's method claims) does not patentably distinguish the prior art — the prior art necessarily achieves the recited result (inherency; In re Hallman, In re Swinehart). Because the pre-1995 art does not do this in a CH3 domain, the correct theory is obviousness, not anticipation — but the functional limitation adds little patentable weight.

---

6. Ground 3 — Domain, product, host-cell and composition claims

These claims rise or fall with Ground 1; they add no independent inventive weight.

---

7. Ground 4 — If the granted set claims the specific mutation pairs

The examples are T366Y (knob) + Y407T (hole) and F405A + T394W, with T366W/Y407A mentioned.

Even the specific pairs are vulnerable, because the prior art identifies the residues: Deisenhofer 1981 and Miller 1990 map the four central anti-parallel β-strands, and the '936's own specification states that "residue T366 is within hydrogen-bonding distance of residue Y407 on the partner CH3 domain … the principal intermolecular contact to residue T366 is to residue Y407 and vice versa." A POSITA starting from the published 2.9 Å structure and told to create a reciprocal knob/hole would naturally start with the closest contact pair on the two central strands — which is exactly T366/Y407. That is "obvious to try" in the KSR sense: a finite set of ~16 candidate residues per domain, with the structurally closest contact pair being the first place to look.

Note, however, a genuine asymmetry: at the 1995 priority date, the T366↔Y407 contact geometry comes from a 2.9 Å model, and the specific volume-matched pair (Y for T at one position, T for Y at the other, "thus maintaining the volume of side chains at the interface") involves a non-obvious reciprocity judgment. This is the single strongest non-obviousness foothold in the claim set — but it is a foothold against specific-pair claims only, and it is defeated if the examiner introduces the Crick/O'Shea 1993 "reciprocal substitution maintaining total volume" teaching.

---

8. Synthesis — would a POSITA have combined these? Explicit motivation analysis

The strongest formulation of the rejection is a single, tightly-linked five-reference combination, not a scattered set:

Chamow 1994 (the co-expression platform and the demonstrated problem) in view of Deisenhofer 1981 and Miller 1990 (the CH3 interface structure and residue map) in view of Crick 1952/53 and O'Shea 1991 (the knob-into-hole packing principle) in view of Ward 1987, John 1994 and O'Shea 1993 (proof that complementary interface engineering flips a homospecific interface to heterospecific).

Every link is explicit:

• Common problem: three of the references (Milstein 1983, Brennan 1985, Chamow 1994) are addressed to the same problem: obtaining a single, correctly paired bispecific/bifunctional heavy-chain species.
• Common field: all are protein-engineering / immunoglobulin-structure references a POSITA would consult together.
• Explicit suggestion: Crick/O'Shea 1993 name the very mechanism ("knobs," "holes," "reciprocal") that the claims recite; Ward 1987 and John 1994 supply the "and you can use this to force heterodimerization" teaching.
• Reasonable expectation of success: Ward 1987 and O'Shea 1993 had already reduced the concept to practice in other systems; the Fc interface is more symmetric and structurally simpler than a coiled coil, so the extension is not speculative.
• Predictable result: a biased hetero-association ratio — which is all the claims require.

Absence of teaching away. Nothing in the retrieved art states that the CH3–CH3 core cannot be engineered, nor that knob-into-hole packing is unsuitable for a β-sandwich.

---

9. Anticipated rebuttals from the patent owner, and how they rate

---

10. Claim-by-claim vulnerability summary

---

11. Practical posture and honest limitations

1. This is largely an academic exercise for enforcement purposes. Per the earlier sections, the patent is Expired – Fee Related with an adjusted expiration of 2016-12-21, no AIA trial was ever filed, and the § 286 six-year lookback from any 2026 complaint reaches back only to ~2020-09 — entirely after expiry. A § 103 holding today would matter chiefly for a stale past-damages defense or for the family's validity narrative.
2. For a continuation, obviousness-type double patenting (ODP) is often the more potent attack than § 103. Because the '936 shares a 1995-03-01 priority and overlapping disclosure with US 5,731,168 / 5,807,706 / 5,821,333 / 7,642,228, a challenger should press ODP over the parent and siblings in parallel with the § 103 grounds above. That is a § 101/§ 103-adjacent judicial doctrine, not § 102/§ 103 art — which is why I flag it separately rather than folding it in.
3. What I could not verify: (a) the verbatim text of the '936's own "Prior Art"/Citations section — the supplied page was truncated before that block; (b) the exact granted 73-claim set; (c) the precise publication date of Belshaw et al. 1995; (d) whether the specific mutation-pair claims survive in the granted set at all.
4. Contradiction carried forward, unresolved: Unified Patents lists the priority date for US-7695936-B2 as 1995-02-28, while Google Patents and the continuity data show 1995-03-01. The one-day difference does not change any conclusion here, but it does shift the § 102(b) bar date by one day (1994-02-28 vs. 1994-03-01), which could matter for a reference published in that window. The cover-page/§ 120 continuity data (1995-03-01) remains the more authoritative.

Bottom line: a strong prima facie § 103 case exists against every generic claim, built entirely on pre-1995 art and turning on the well-characterized CH3 interface (Deisenhofer/Miller), the literal knob-into-hole packing principle (Crick/O'Shea), the demonstrated conversion of a symmetric interface into a heterospecific one (Ward/John/O'Shea 1993), and an art-recognized production problem (Milstein/Cuomo–Chamow). The patent's defensible ground is narrow and resides, if anywhere, in the specific reciprocal, volume-matched mutation pairs — not in the genus.