[A retrieval method of drug molecules based on graph collapsing].

Area of Science:

• Computational Chemistry
• Bioinformatics
• Information Retrieval

Background:

• Chemical structural formulas (CSFs) are precise molecular identifiers crucial for medicine information retrieval.
• Existing graph-based retrieval systems face challenges in efficiency and accuracy.
• Developing compact and efficient molecular representations is key for advanced search functionalities.

Purpose of the Study:

• To establish a compact and efficient hypergraph representation for molecules.
• To develop a graph-similarity-based retrieval method for effective medicine information retrieval.
• To improve the accuracy and efficiency of searching chemical structures.

Main Methods:

• A novel hypergraph representation for molecules was developed, analyzing factors affecting graph matching efficiency.
• A hierarchical collapsing method combining graph isomorphism and frequent subgraph mining was employed.
• A graph-isomorphism-based algorithm was proposed to address subgraph overlapping during hypergraph construction.
• Multi-dimensional measures were used to evaluate spatial similarity among graphical CSFs.

Main Results:

• The proposed system demonstrated superior retrieval accuracy compared to the state-of-the-art Wikipedia Chemical Structure Explorer (WCSE).
• Key performance metrics including mean average precision and rank-biased precision showed significant improvements over WCSE.
• The system achieved higher retrieval accuracy for top-10 results, indicating enhanced effectiveness.

Conclusions:

• A novel graph-similarity-based retrieval approach for medicine information has been proposed.
• An effective and efficient hypergraph representation and an isomorphism-based algorithm for dominant subgraph selection were developed.
• Experimental results validate the proposed approach's effectiveness and superiority in medicine information retrieval.