Application of Multiblock Analysis on Small Metabolomic Multi-Tissue Dataset.

Area of Science:

• Metabolomics
• Bioinformatics
• Data Integration

Background:

• Data integration offers valuable insights by revealing common and unique trends across different data blocks.
• Analyzing small multiblock datasets is challenging due to reduced statistical power and limited method choices.
• Metabolomic studies often involve multiple tissue types, necessitating integrated analysis.

Purpose of the Study:

• To investigate the utility of multiblock analysis for small metabolomic datasets with limited samples.
• To compare the joint and unique multiblock analysis (JUMBA) with concatenated principal component analysis (PCA) for trend detection.
• To identify underlying factors contributing to metabolic variations in a small multiblock metabolomic dataset.

Main Methods:

• Studied a small metabolomic multiblock dataset comprising six tissue types with common metabolites.
• Applied the joint and unique multiblock analysis (JUMBA), a single model method.
• Compared JUMBA's performance against concatenated principal component analysis (PCA).

Main Results:

• JUMBA successfully detected trends and identified underlying factors responsible for metabolic variations.
• Extracted components using JUMBA were interpretable and linked to relevant biological properties.
• JUMBA elucidated relationships between observations, their stability, and block contributions to components.

Conclusions:

• Multiblock analysis methods, such as JUMBA, are valuable for small metabolomic datasets.
• JUMBA demonstrates effectiveness in identifying metabolic variations and biological relevance even with limited sample sizes.
• The study supports the application of multiblock analysis in scenarios with low statistical power.