Predicting pathways for old and new metabolites through clustering.

Area of Science:

• Biochemistry
• Metabolomics
• Bioinformatics

Background:

• Metabolic pathways are crucial for life, but identifying them for novel metabolites is difficult.
• Currently, only 60% of metabolites in the Human Metabolome Database (HMDB) are assigned to known pathways.
• Elucidating metabolic pathways for new compounds is time-consuming and resource-intensive.

Purpose of the Study:

• To develop and validate a computational approach for predicting metabolic pathways based on metabolite structure.
• To improve the assignment rate of metabolites to their respective pathways.
• To facilitate the discovery and understanding of novel metabolic functions.

Main Methods:

• Extraction of 201 structural features from SMILES (Simplified Molecular Input Line Entry System) annotations.
• Identification of new metabolites from PubMed abstracts and the HMDB.
• Application of clustering algorithms to metabolite features to identify correlations.
• Quantification of correlations between metabolites to link them to pathways.

Main Results:

• The developed approach successfully linked 92% of known metabolites to their correct metabolic pathways.
• Clustering algorithms effectively identified relationships between metabolite structures and their functions.
• The method demonstrated high accuracy in pathway prediction for known metabolites.

Conclusions:

• This structure-based computational approach offers a valuable tool for predicting metabolic pathways of newly discovered metabolites.
• The findings significantly advance the field of metabolomics by addressing the challenge of pathway annotation.
• The method has the potential to accelerate biological research by enabling faster functional characterization of metabolites.