Optical Chemical Structure Recognition turns structure images into machine-readable chemistry.
i Evaluated across 420+ randomly selected patent documents, 8.8K+ structures and 5.1K+ complete molecules.
97.06%OCSR precision
89.74%OCSR F1 score
AI agents for life sciences
Turn hidden patent data into smarter chemistry decisions.
Lead Compound Analyzer finds the compounds most likely to be the lead. SAR Extractor turns the structures and activity values into one reviewable table.
From disclosure to chemistry insight
Trusted by over 1,000 life science innovators.
Focus your compound review
Lead Compound AnalyzerBeta
Reads the biological and experimental evidence disclosed in a patent and shortlists the likely leads, with the rationale for each.
- Identify up to 10 likely lead compounds from a single patent.
- Review potency, selectivity, in vivo findings and developability context where available.
- Compare candidates grouped by evidence level, with supporting rationale and references.
Bring the findings to your team Word · Excel · SDF for recommended molecules
Lead Compound Analyzer
Input WO2022232536A1 One small molecule patent
Reading the evidence Potency · selectivity · in vivo · developability
Molecule recommendation Evidence level shown · not ranked
99 
IKZF2 DC₅₀ 0.001 µM · 89% reduction at 30 µM, 6 hStructure P57 · Activity P98 Level 3
IKZF2 DC₅₀ 0.001 µM · 89% reduction at 30 µM, 6 hStructure P57 · Activity P98 Level 3 72 IKZF2 DC₅₀ 0.001 µM · 80% reduction at 30 µM, 6 hStructure P53 · Activity P98 Level 3
IKZF2 DC₅₀ 0.001 µM · 80% reduction at 30 µM, 6 hStructure P53 · Activity P98 Level 3 111 IKZF2 DC₅₀ 0.005 µM · hERG IC₅₀ >30 µMStructure P59 · Activity P98 Level 3
IKZF2 DC₅₀ 0.005 µM · hERG IC₅₀ >30 µMStructure P59 · Activity P98 Level 3 + 7 more candidates in this report, each with rationale, references and predicted ADMET
Connect structures to activity
SAR ExtractorBeta
Uncover structure–activity relationships across patent disclosures without reconstructing every table by hand.
- Process up to 10 small molecule patents per task, depending on your plan.
- Connect structures with activity values, targets and experimental context.
- Review confidence scores, run in-platform SAR analysis and export results for further investigation.
Take the data into your workflow CSV and Excel activity data · SDF structures and SMILES
SAR Extractor
WO2026132447A1Activity details · 311Compound summary · 274
Cpd 1 MTHFD2 · IC₅₀50–500 nM HighHighTable 2 · P30
Cpd 2 MTHFD2 · IC₅₀500–5000 nM HighHighTable 2 · P30
Cpd 5 MTHFD2 · IC₅₀<50 nM HighHighTable 2 · P30
Confidence is scored separately for structure and activity. Three of 311 extracted activity rows shown.
01Extract structure & activity 02Review confidence & sources 03In-platform SAR analysis 04Export for further analysis
Near-real-time patent intelligence. Built for scientific review.
Patsnap’s OCSR, NER and NOR models connect chemical drawings, names and compound identities with patent evidence. Combined with layout analysis and ID mapping, they help turn new disclosures into usable chemistry data without waiting for manual curation.
Named Entity Recognition identifies chemical entities in text and supports structure recovery from chemical names.
Entity normalization connects names and compound codes with known drug records where applicable.
Three ways teams use SAR Extractor and Lead Compound Analyzer.
Guide your next compound design
Compare structures and reported activity to inform hit-to-lead and lead optimization hypotheses.
Understand emerging competitor chemistry
Investigate newly disclosed compound series and likely leads alongside your program priorities.
Build reusable research datasets
Export structured records for expert curation, cheminformatics analysis and downstream modeling.
Frequently asked questions.
Which patents are best suited to these agents?
SAR Extractor focuses on organic small molecule patents with quantitative activity data and molecules below 900 Da. Files can be up to 50 MB each. Quality may decline for documents above 500 pages. Markush structures, biologic macromolecules, polymers and formulation-only patents are outside its scope.
Lead Compound Analyzer primarily supports small molecule patents and currently accepts a single patent number. Process, synthesis, formulation and method patents are not supported for lead analysis.
Can I analyze a newly published patent?
SAR Extractor supports direct PDF upload, so you can analyze a newly published document before it becomes available through patent-number search, within supported file and document limits.
Do the outputs still need scientific review?
Yes. Both agents are in beta. Extraction is not exhaustive, and performance varies with document length, layout and chemistry complexity. Use confidence scores, source references and expert review to validate critical findings before reuse.
How does Lead Compound Analyzer prioritize compounds?
It analyzes the biological and experimental evidence disclosed in a patent to identify the compounds most likely to be the lead. Up to 10 candidates are grouped by supporting evidence level for expert assessment, rather than assigned a numerical rank. Predicted ADMET properties are distinguished from experimental findings.
Can I export results to my existing tools?
SAR Extractor supports CSV and Excel for activity data and SDF for structures and SMILES. Lead Compound Analyzer supports Word, Excel and SDF exports for recommended molecules. Usage limits depend on your subscription.
