PatentNetML: A Novel Framework for Predicting Key Compounds in Patents Using Network Science and Machine Learning

Area of Science:

• Drug discovery and development
• Computational chemistry
• Intellectual property analysis

Background:

• Patents are vital for drug research, offering early data and insights.
• Identifying key compounds in patents is crucial for discovering novel lead compounds.
• Existing methods may not fully leverage the information within patent data.

Purpose of the Study:

• To develop an innovative approach for predicting key compounds within patents.
• To create a robust framework integrating network science and machine learning for this purpose.
• To demonstrate the utility of the proposed framework through case studies.

Main Methods:

• Collected a dataset of 1555 patents and 1000 key compounds.
• Developed the PatentNetML framework, integrating network science and machine learning algorithms.
• Combined network measures, ADMET properties, and physicochemical properties for classification models.

Main Results:

• Successfully constructed classification models to identify key compounds.
• Demonstrated the potential of PatentNetML in uncovering hidden patterns in patents.
• Showcased the framework's capability through model interpretation and case study analysis.

Conclusions:

• PatentNetML offers a promising foundation for efficiently identifying drug candidates.
• The framework aids in expediting the drug discovery process in the pharmaceutical industry.
• Acknowledged limitations exist for patents deviating from the assumed central pattern.