Core trees and consensus fragment sequences for molecular representation and similarity analysis

Area of Science:

• Computational chemistry and cheminformatics.
• Molecular modeling and drug discovery.

Background:

• Traditional molecular representations face challenges in capturing complex structure-activity relationships.
• Developing efficient methods for analyzing and comparing active molecules is crucial for drug discovery.

Purpose of the Study:

• To introduce a new molecular representation based on activity class characteristic substructures.
• To enable efficient mapping and database searching of active compounds.

Main Methods:

• Extracting characteristic substructures from random fragment populations.
• Mapping substructures to determine atom match rates in active molecules.
• Defining a hierarchical molecular fragmentation scheme based on atom match rates.
• Encoding active compounds as fragmentation pathways from core trees.
• Applying biological sequence alignment methods and substructure-based scoring functions.
• Deriving consensus fragment sequences from multiple core path alignments to represent activity classes.

Main Results:

• A novel hierarchical molecular fragmentation scheme was established.
• Active compounds were successfully encoded as fragmentation pathways.
• Consensus fragment sequences were derived to represent compound activity classes.
• The method demonstrated potential for mapping molecules and searching databases for active compounds.

Conclusions:

• The new molecular representation effectively captures structure-activity relationships.
• This approach offers a powerful tool for analyzing compound activity classes.
• The method enhances the efficiency of searching large chemical databases for potential drug candidates.