Mining patents using molecular similarity search

Area of Science:

• Biomedical research
• Cheminformatics
• Intellectual Property (IP) analysis

Background:

• Text analytics is crucial in biomedical research.
• Core algorithms for entity identification and relation extraction are advancing.
• Practical applications for text analytics in this field are needed.

Purpose of the Study:

• To develop a system for exploring the US Patent corpus using molecular information.
• To provide a practical application for text analytics in biomedical research and IP discovery.
• To improve the identification of intellectual property compared to traditional methods.

Main Methods:

• Developed a system integrating a chemical annotator, a similarity search engine using the International Chemical Identifier (InChI) standard, and data mining tools.
• Indexed 3,623,248 unique chemical structures from 4,375,036 US Patents and Patent Applications.
• Enabled users to draw molecules, search for related IP, and analyze results via a web interface.

Main Results:

• Successfully identified and indexed a large number of unique chemical structures from the US patent corpus.
• Demonstrated the system's capability to facilitate IP exploration based on molecular information.
• Showcased a more effective approach to IP identification than conventional keyword-based searches.

Conclusions:

• The developed system offers a powerful new method for exploring the US Patent corpus using molecular data.
• This approach significantly enhances the discovery and analysis of intellectual property.
• Leveraging chemical information in patent searches provides superior results compared to keyword-based strategies.