Predicting key example compounds in competitors' patent applications using structural information alone

Area of Science:

• Drug Discovery
• Medicinal Chemistry
• Intellectual Property Analysis

Background:

• Predicting key compounds in competitor patents is vital for drug discovery programs.
• Current methods rely on limited information like biological data or reaction scale, which is often insufficient.
• Medicinal chemists traditionally identify key compounds through extensive structure-activity relationship (SAR) studies.

Purpose of the Study:

• To develop a novel method for predicting key compounds in competitor patent applications using only structural information.
• To provide an alternative approach when traditional methods are hindered by a lack of data.
• To complement existing strategies for patent analysis in drug discovery.

Main Methods:

• The study utilized a computational approach based on chemical structure analysis.
• Compounds were represented using Extended Connectivity Fingerprints (ECFPs) to map chemical space.
• The method identifies compounds situated in densely populated regions within the chemical space of patent examples, assuming these are central to SAR studies.

Main Results:

• The method was validated on 30 patents containing launched drugs.
• The approach successfully predicted the key compounds (launched drugs) in 17 out of 30 patents (57% success rate).
• This demonstrates the potential of structure-based prediction in identifying critical molecules within patent literature.

Conclusions:

• The developed method offers a viable alternative for predicting key compounds, especially when traditional information is scarce.
• This computational approach can serve as a valuable tool to aid medicinal chemists in patent analysis.
• The findings suggest that focusing on chemical space density can effectively pinpoint significant compounds in drug discovery patents.