BioGrade · expert reasoning data for biopharma agentsEnvironmentsBenchmarkCompare modelsData flywheelThesis
← Patent / Freedom-to-Operate Analysis comparison · model: GPT-4o mini
LLM-judge score: 2.0 / 5 · flawed
search_completeness: 2 claim_overlap: 2 status_expiry: 2 verdict_designaround: 2 faithfulness: 2
Despite running multiple searches, the agent surfaced only one family (US11078291B2) and missed the other two blocking granted composition families (LaNova WO2019173420A1, Ganymed US9051370B2) plus the pending WATCH families. The Sotio claim text is described as a non-cleavable thioether/maytansinoid construct, leading the agent to wrongly suggest the proposed MMAE/cleavable-linker product designs around it — when the broader corpus reads on the product and the correct verdict is BLOCKED. No expired-family handling, no overall verdict, and design-arounds are generic and untethered to the actual blocking scope.
BD / IP · IP / patent analyst

Patent / Freedom-to-Operate Analysis — agent trajectory for expert review

gpt-4o-mini · 2026-06-06T10:23:54.956732+00:00 · 6 tool calls · 11.5s · 4928 tokens
run id: fto_patent_20260606_031816_fto-patent__gpt-4o-mini

Query

Assess freedom-to-operate for a proposed anti-CLDN18.2 antibody-drug conjugate (MMAE auristatin payload, protease-cleavable peptide linker) for gastric cancer in the US. Identify any blocking claims and credible design-arounds.

Trajectory

Agent · turn 1
⚙ calls search_patents(target=antibody-drug conjugate, modality=MMAE, keyword=CLDN18.2, jurisdiction=US)
↳ search_patents result
show data
{
  "count": 0,
  "families": []
}
Agent · turn 2
⚙ calls search_patents(target=anti-CLDN18.2, modality=antibody-drug conjugate, jurisdiction=US)
↳ search_patents result
show data
{
  "count": 0,
  "families": []
}
Agent · turn 3
⚙ calls search_patents(target=CLDN18.2, modality=antibody-drug conjugate, jurisdiction=US)
↳ search_patents result
show data
{
  "count": 1,
  "families": [
    {
      "publication_number": "US11078291B2",
      "jurisdiction": "US",
      "assignee": "Sotio Biotech",
      "title": "Maytansinoid antibody-drug conjugates targeting CLDN18.2",
      "target": "CLDN18.2",
      "modality": "Antibody-drug conjugate",
      "priority_date": "2017-06-15",
      "expiry": "2038-06-15",
      "legal_status": "GRANTED",
      "claim_types": [
        "composition"
      ],
      "keywords": [
        "CLDN18.2",
        "ADC",
        "maytansinoid",
        "DM1",
        "DM4",
        "non-cleavable linker",
        "composition of matter"
      ]
    }
  ]
}
Agent · turn 4
⚙ calls get_claims(publication_number=US11078291B2)
↳ get_claims result
show data
{
  "publication_number": "US11078291B2",
  "title": "Maytansinoid antibody-drug conjugates targeting CLDN18.2",
  "legal_status": "GRANTED",
  "independent_claims": [
    {
      "claim_no": 1,
      "claim_type": "composition",
      "text": "An antibody-drug conjugate comprising an anti-CLDN18.2 antibody, a non-cleavable thioether linker, and a maytansinoid (DM1/DM4) payload."
    }
  ]
}
Agent · turn 5
⚙ calls search_patents(target=antibody-drug conjugate, modality=MMAE, jurisdiction=US)
↳ search_patents result
show data
{
  "count": 0,
  "families": []
}
Agent · turn 6
⚙ calls search_patents(target=MMAE, modality=antibody-drug conjugate, jurisdiction=US)
↳ search_patents result
show data
{
  "count": 0,
  "families": []
}
Agent · turn 7

Freedom-to-Operate Assessment for Anti-CLDN18.2 Antibody-Drug Conjugate

Based on the search conducted, the following significant findings have been made regarding the proposed anti-CLDN18.2 antibody-drug conjugate with MMAE (auristatin payload) and a protease-cleavable peptide linker:

Identified Patent:

Blocking Claims

The identified patent contains blocking claims relevant to your proposed product:
- The claim encompasses an antibody-drug conjugate that includes an anti-CLDN18.2 antibody and a maytansinoid payload. This could potentially block any similar ADC targeting CLDN18.2 that does not sufficiently differentiate from the claimed invention.

Design-Around Strategies

  1. Linker Modification:
    - Since the blocking claim specifies a "non-cleavable thioether linker", using a protease-cleavable peptide linker could serve as a distinction to navigate around the claim.

  2. Payload Alteration:
    - Rather than using a maytansinoid payload (DM1/DM4), alternative payloads like auristatins (MMAE) may provide a credible design-around, as long as the other components of the ADC (antibody and linker) are also modified sufficiently.

  3. Antibody Variants:
    - Development of a different anti-CLDN18.2 antibody or different antibody variants that target a unique epitope on the CLDN18.2 antigen could provide a route to differentiation.

Conclusion

Given the grant status and the composition claim related to anti-CLDN18.2 ADCs, there may be substantial limitations for the proposed MMAE payload ADC unless alternative designs are pursued. Careful consideration of the claim scope and strategic alterations can facilitate a pathway for action in this domain.

Final output

Freedom-to-Operate Assessment for Anti-CLDN18.2 Antibody-Drug Conjugate

Based on the search conducted, the following significant findings have been made regarding the proposed anti-CLDN18.2 antibody-drug conjugate with MMAE (auristatin payload) and a protease-cleavable peptide linker:

Identified Patent:

Blocking Claims

The identified patent contains blocking claims relevant to your proposed product:
- The claim encompasses an antibody-drug conjugate that includes an anti-CLDN18.2 antibody and a maytansinoid payload. This could potentially block any similar ADC targeting CLDN18.2 that does not sufficiently differentiate from the claimed invention.

Design-Around Strategies

  1. Linker Modification:
    - Since the blocking claim specifies a "non-cleavable thioether linker", using a protease-cleavable peptide linker could serve as a distinction to navigate around the claim.

  2. Payload Alteration:
    - Rather than using a maytansinoid payload (DM1/DM4), alternative payloads like auristatins (MMAE) may provide a credible design-around, as long as the other components of the ADC (antibody and linker) are also modified sufficiently.

  3. Antibody Variants:
    - Development of a different anti-CLDN18.2 antibody or different antibody variants that target a unique epitope on the CLDN18.2 antigen could provide a route to differentiation.

Conclusion

Given the grant status and the composition claim related to anti-CLDN18.2 ADCs, there may be substantial limitations for the proposed MMAE payload ADC unless alternative designs are pursued. Careful consideration of the claim scope and strategic alterations can facilitate a pathway for action in this domain.

Reference answer key (deterministic) — open after grading

Heuristic reference over the curated corpus for target=CLDN18.2, modality=Antibody-drug conjugate. Families in scope: 6 ({'GRANTED': 3, 'PENDING': 2, 'EXPIRED': 1}). BLOCKING = granted composition-of-matter on the same target+modality and not expired; WATCH = pending or method-only; CLEARABLE = expired. This is illustrative reasoning over a teaching snapshot, NOT legal advice.

Publication #AssigneeLegal statusExpiryClaim type(s)FTO relevance
WO2019173420A1LaNova MedicinesGRANTED2039-03-08composition, methodBLOCKING (granted composition-of-matter)
CN114106192AKeymed BiosciencesPENDING2040-09-30compositionWATCH (pending — claims not yet fixed)
US11078291B2Sotio BiotechGRANTED2038-06-15compositionBLOCKING (granted composition-of-matter)
US9051370B2Ganymed Pharmaceuticals / AstellasGRANTED2029-01-03composition, methodBLOCKING (granted composition-of-matter)
WO2016166122A1Astellas PharmaPENDING2036-04-13methodWATCH (pending — claims not yet fixed)
CN103509118AGanymed PharmaceuticalsEXPIRED2018-11-24compositionCLEARABLE (expired)
OVERALL FTO VERDICT (heuristic)BLOCKED — 3 granted composition-of-matter family(ies) read on the proposed product; need a licence or a design-around.

Expert grading

Corpus search completeness 1 (poor) 5 (excellent)
5: Systematically searches the corpus (orients via list_corpus, then queries by target, modality, keyword, and across CN/US/WO) and surfaces the full set of relevant families for the proposed product — including the same-target naked-antibody composition family that can read on an ADC built from it.
1: Runs one narrow query or reasons from memory; misses blocking families or whole jurisdictions; treats the corpus as smaller than it is.
Claim-overlap analysis 1 (poor) 5 (excellent)
5: Reads actual claim text and distinguishes composition-of-matter (molecule/ADC/sequence) from method-of-treatment, and genus from species — e.g. recognises an auristatin/MMAE genus ADC claim reads on a specific MMAE construct, or that a naked-antibody composition claim covers the antibody used in an ADC.
1: Conflates a narrow method claim with a composition claim, ignores genus/species scope, or asserts overlap/non-overlap without reading the claim text.
Legal-status & expiry handling 1 (poor) 5 (excellent)
5: Correctly treats expired families as non-blocking, separates granted from pending (pending claims are not yet enforceable but are a WATCH), and uses priority/expiry dates to reason about when a blocker falls away.
1: Calls an expired patent blocking, treats a pending application as a hard block (or ignores it entirely), or never checks legal status / expiry dates.
FTO verdict & design-around quality 1 (poor) 5 (excellent)
5: Renders a CLEAR/WATCH/BLOCKED verdict that follows from the claims found, and proposes credible, specific design-arounds (e.g. switch from a claimed MMAE/auristatin payload to an unclaimed payload class, change linker chemistry, or wait out an expiring family) tied to the actual blocking claim scope.
1: Verdict contradicts the evidence, is hedged into uselessness, or design-arounds are generic ('use a different antibody') and not tied to specific blocking claims.
Evidence faithfulness 1 (poor) 5 (excellent)
5: Every cited family, claim number, assignee, status, and date traces to tool output; claim-scope statements quote or paraphrase the real claim text; no invented patents.
1: Fabricates publication numbers, claims, assignees, or dates, or contradicts the returned corpus data.
Overall verdict
Your name / ID