High-Quality Predicted Pathway Annotations Greatly Improve Pathway Enrichment Analysis of Metabolomics Datasets

Area of Science:

• Computational Biology
• Metabolomics
• Bioinformatics

Background:

• Metabolomics data interpretation relies on aggregating metabolite information, often using pathway enrichment analysis (PEA).
• Accurate metabolite-pathway annotations are crucial for effective PEA, but current knowledgebases have incomplete annotations.
• This incompleteness limits the biological insights derived from PEA.

Purpose of the Study:

• To develop a novel method for generating high-quality metabolite-pathway annotations.
• To enhance the effectiveness of pathway enrichment analysis (PEA) in metabolomics.
• To improve the biological and biomedical interpretability of metabolomics datasets.

Main Methods:

• Utilized a novel multitask classification, graph convolutional-like neural network.
• Generated high-quality metabolite-pathway annotations across KEGG, MetaCyc, and Reactome databases.
• Applied predicted annotations to 990 Metabolomics Workbench datasets for PEA.

Main Results:

• Achieved a significant improvement in metabolite-pathway annotation quality.
• Demonstrated over a 10-fold increase in the median number of enriched pathways detected.
• Enhanced datasets showed substantially improved biological and biomedical interpretability.

Conclusions:

• The novel neural network approach effectively generates high-quality metabolite-pathway annotations.
• Improved annotations substantially increase the yield of PEA, enhancing data interpretation.
• This method offers a powerful tool for advancing metabolomics research and discovery.