Protein interaction network constructing based on text mining and reinforcement learning with application to prostate...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Network Science

Background:

• Constructing protein interaction networks from biomedical literature is crucial for understanding biological systems.
• Existing methods like co-occurrence and linguistic patterns have limitations in efficiency and precision.

Purpose of the Study:

• To develop an integrated approach for extracting protein interactions from biomedical texts.
• To establish a robust protein interaction network using reinforcement learning.
• To validate the constructed network's properties against known biological network characteristics.

Main Methods:

• A hybrid algorithm combining linguistic patterns and co-occurrence approaches for interaction extraction.
• A reinforcement learning-based agent to build and optimize the protein interaction network.
• Utilizing PubMed-downloaded texts for constructing a prostate cancer-specific network.

Main Results:

• The proposed method achieved a good matching rate in extracting protein interactions.
• The constructed network's topology, including node degree distribution, aligns well with scale-free network properties.
• Demonstrated effective construction of a prostate cancer protein interaction network.

Conclusions:

• The integrated text mining and reinforcement learning approach is effective for building biological networks.
• The developed method offers a promising strategy for analyzing protein interactions in specific disease contexts.
• The generated networks exhibit characteristics consistent with biological systems, specifically scale-free topology.