Prior art searching · Chemical structures
The prior art that keywords will never return.
In chemical patents, the most dangerous disclosure is usually one nobody named. It sits inside a generic claim — drawn once, never written out. Structure search is how you find it.
One drawing. Thousands of disclosures. No names.
A keyword prior art search assumes the invention was described in words. In chemistry, it usually wasn’t. It was drawn as a generic structure with placeholders, and the claim text quietly defines every variation it covers.
That’s a Markush structure: a fixed core scaffold carrying labelled positions — R1, R2, X — plus a list of what each is allowed to be. Nothing in that document holds your compound’s name, its registry number, or a drawing of it specifically. Your compound may still be squarely inside it.
|
R3 — CORE SCAFFOLD — R2
fixed skeleton · variable positions
…a compound of Formula I, wherein R1 is selected from CH3, CF3, C2H5, cyclopropyl; R2 is selected from Cl, F, OCH3, OH, H; R3 is selected from H, NH2, CH3.
4 × 5 × 3 = 60 compounds disclosed by one drawing.
Every one is potentially citable prior art. None is searchable by name, and only a handful will ever appear as a worked example. This is the gap structure search exists to close.
Four structure search types, four different answers
“Structure search” isn’t one thing. It’s four — and running the wrong one is how a novelty search comes back clean on a compound disclosed a decade ago.
| Search type | The question it answers | Blind to | Recall | Effort |
|---|---|---|---|---|
| Exact structure | Has this precise compound been disclosed before? | Everything adjacent — one methyl group defeats it | Narrow | Low |
| Substructure | Who has disclosed anything containing our core? | Analogs where the core itself was altered | High | Low |
| Markush / claim scope | Does our compound fall inside somebody’s generic claim? | Anything outside the claim’s defined lists, however close | Scoped | High |
| Similarity | What else is chemically near this, on any scaffold? | Little — the trade is noise, which scales with reach | Broad | Medium |
Recall and effort are directional planning guidance, not measured benchmarks.
Six prior art records. Four verdicts.
One query compound, one small prior art set. Switch the search type and watch which records surface.
Note PA-02. An exact search says the compound is new. A claim-scope search finds a pending application that already covers it. That difference separates a filing decision from a clearance problem.
Which search does your decision need?
The search type follows the business question, not the chemistry. Pick your mandate.
Exact and substructure, backed by claim-scope
Establish quickly whether the compound has been disclosed outright, then widen to substructure. The step teams skip is claim-scope searching — a generic claim can anticipate a compound nobody ever named.
Deliverable · novelty report with ranked art and anticipation notes
Markush claim-scope, with legal status screening
Freedom to operate turns on whether a live claim covers what you intend to sell, so resemblance isn’t the test — membership is. We search generic claims directly, then filter to your markets and check status.
Deliverable · FTO report with claim-by-claim analysis and jurisdiction status
Substructure and similarity against the priority date
You need art that predates their filing and reads on their claims. Broad structural coverage plus a similarity sweep finds the disclosures their examiner didn’t, including non-patent literature.
Deliverable · invalidation search with claim charts mapping art to claim elements
Substructure and similarity, analysed by assignee and date
Landscaping rewards recall over precision. Cast wide across the scaffold space, then let the filing pattern show who is working where and which chemistry is still open.
Deliverable · landscape report with scaffold clustering and white-space analysis
The volume behind a single claim
Generic claims scale multiplicatively. A modest one covers more compounds than any team could review by hand — which is why claim-scope searching is indexed and tested, not read.
Counts assume every listed combination is legal. Real claims are smaller — valence rules, mutually exclusive definitions and proviso language prune the total. They rarely prune it to a size you can eyeball.
Where chemical prior art actually sits
A structure search is only as good as the sources it runs against. Patent chemistry is spread across offices, registries and literature that index structures differently — or not at all.
Well indexed by structure
- Granted patents, major offices
- Published applications, PCT
- Chemical registries
- Commercial compound databases
Partial — needs expert handling
- Markush claims in full text
- Non-Latin-script filings
- Older, image-only records
- Sequence-adjacent chemistry
Text-only — structure search won’t reach it
- Journal articles, abstracts
- Theses, posters
- Regulatory filings
- Product literature
This is why a structure search alone isn’t a prior art search. Structure finds the chemistry; targeted text and classification searching finds the rest.
How a Synoptic IP structure search runs
Each stage narrows what the next has to handle, and every stage is recorded so the search is reproducible if it’s ever challenged.
-
Brief and scope
We agree the decision the search supports, the jurisdictions that matter, the cut-off date, and how much recall you’ll trade for a shorter review list. This conversation shapes the result more than any tool does.
-
Query construction
We build the query core at the smallest size that still describes the invention. Every atom added to a query is a hit you’ll never see — so this step is deliberate, documented, and iterated with your chemists.
-
Multi-database execution
Structure queries across patent and registry sources, run in parallel with classification and text searching for disclosure that was never indexed structurally.
-
Expert screening
Raw hits get read, not just counted. We check whether a generic claim genuinely reads on your compound, whether the provisos exclude it, and whether the record is live in the markets you care about.
-
Reporting and handover
You receive ranked, annotated art with the reasoning attached — plus the full query log, so your attorneys can reproduce or extend the search rather than start again.
- Ranked artRecords ordered by how closely they read on the query, with the specific claim or passage identified.
- Relevance notesWhy each record matters — anticipation, obviousness combination, or claim-scope overlap — in language your team can act on.
- Structure evidenceQuery structure alongside each hit’s disclosed structure or generic claim, so relevance is visible rather than asserted.
- Legal statusGrant status, family members and jurisdictions for anything bearing on freedom to operate.
- Search logDatabases, query strings, structure files, filters and dates — the audit trail that makes the search defensible.
What to send with your brief
Searches stall for the same handful of reasons. Tick these off and we can start on day one.
What clients ask us
Can’t our team just run this in a free database?
You can run exact and substructure queries in public tools, and for a quick directional read that’s often enough. What public tools don’t do well is claim-scope searching across generic claims, coverage of offices with limited structure indexing, and the expert screening that decides whether a generic claim actually reads on your compound. That’s where searches come back falsely clean.
Is a substructure hit the same as an infringement finding?
No. A hit means a record contains your fragment. Whether a claim covers your compound depends on the defined substituent lists read together with the provisos — and whether that amounts to infringement is a legal question for your counsel. We identify and evidence the art; your attorneys draw the conclusion.
Do you search non-patent literature too?
Yes, and it’s necessary. A significant share of chemical disclosure appears first in journals, theses, conference material and regulatory filings that are not structure-indexed. We pair structure queries with targeted text and classification searching to reach it.
What if we don’t have a single defined compound yet?
Common in early-stage programmes, and workable. We search the genus rather than a point compound — the core plus the substituent ranges you’re considering — which also shows which parts of your intended claim are already occupied.
Can you handle filings in Chinese, Japanese or Korean?
Yes. These are exactly the filings where structure searching earns its keep, because text searching across scripts and transliterated chemical names is unreliable. Structures are language-independent; we search them as structures and confirm relevance from the original document.
How do you keep our chemistry confidential?
Query structures for unfiled inventions are handled under NDA and searched in environments appropriate to their sensitivity. Tell us at briefing if the structure is unpublished and we’ll confirm arrangements before anything is run.
Bring us the structure. We’ll bring back the art.
Novelty, freedom to operate, invalidation or landscape — for pharma and biotech teams whose decisions turn on chemistry that keyword searching can’t see.
Educational content — not legal advice