Patent Extractor

The Antibody Patent Extractor, developed by Converge Bio, is a tool designed for drug discovery teams that need to extract structured antibody data from patents quickly and accurately. It processes US, EP, AU, and WO patent documents to recover every antibody sequence, complementarity-determining region (CDR), and affinity measurement contained within them, returning results as structured data typically within two hours rather than the three to five days required for manual extraction.

The tool is aimed at discovery teams performing patent landscape analysis, competitive intelligence, or sequence mining as part of antibody research and development workflows. It requires no account or credit card to try, accepting a patent ID as input and delivering results via email along with a downloadable file and a shareable viewer link.

Core Extraction Capabilities

• Accepts patent identifiers from US, EP, AU, and WO patent systems
• Retrieves the full patent PDF, runs optical character recognition (OCR), and constructs a complete evidence graph automatically
• Extracts every antibody sequence, CDR annotation, and affinity measurement present in the document
• Assigns confidence scores to extracted data points to help separate signal from noise
• Links every claim back to its source sentence or table within the patent
• Captures full kinetics, selectivity, stability, in vivo data, epitope information, and additional experimental measurements where present

Output and Delivery

• Results are delivered by email upon completion, typically within approximately two hours
• Output is provided as an XLSX file containing every antibody, CDR, affinity measurement, and confidence score
• A shareable viewer link is also generated alongside the downloadable file
• Data is structured into a graph format that organizes extracted evidence systematically

Performance Benchmarks

• Recovers 2.8 times more antibodies compared to PLAbDab on the same patent set
• Achieves 98% agreement with PLAbDab on overlapping data, indicating high extraction accuracy
• Reduces typical turnaround from three to five days of manual work to approximately two hours

Integration with Downstream Design Workflows

• Extracted sequences can serve as seed inputs for antibody design passes using Converge Bio's ConvergeAB platform
• A published case study demonstrates extraction of the cetuximab sequence from patent US20160002330, which was then used in a zero-shot design pass with no additional experimental data
• The resulting Converge variant achieved a KD of 315 pM, representing 2.1 times higher affinity than cetuximab, with only six sequence edits from the parent drug

The tool is positioned as a starting point for antibody discovery workflows, enabling teams to move from patent literature to structured, actionable sequence data rapidly, and optionally feeding those sequences directly into computational antibody optimization pipelines.