Local neighbor Normalization: Reconciling accurate normalization and heterogeneity recovery in large-scale...

Area of Science:

• Metabolomics
• Bioinformatics
• Data Analysis

Background:

• Metabolomics studies face challenges with sample concentration variations (dilution effect) impacting data interpretation.
• Traditional normalization methods (CSN, PQN, MDFC) often overlook data heterogeneity, potentially masking true biological differences.
• This heterogeneity is crucial for distinguishing biological subgroups in complex datasets.

Purpose of the Study:

• To introduce Local Neighbor Normalization (LNN), a novel method for correcting dilution effects in metabolomics.
• To preserve intrinsic data variability and local data structures during normalization.
• To improve the accuracy of downstream analyses, including biomarker discovery.

Main Methods:

• LNN normalizes samples using a tailored reference spectrum derived from a local neighbor set identified by similarity metrics.
• The method constructs sample-specific reference spectra, accounting for local data structure.
• Evaluations were performed on simulated and real metabolomics datasets (NMR, GC-MS, LC-MS).

Main Results:

• LNN demonstrated superior performance in eliminating dilution effects compared to CSN, PQN, and MDFC.
• The method effectively recovered inter-sample heterogeneity and inter-metabolite correlations.
• LNN showed particular strength in datasets with over 50% differential metabolites, preserving local data structures.

Conclusions:

• LNN offers a robust approach to metabolomics data normalization, addressing limitations of existing methods.
• Its ability to preserve data heterogeneity and local structures is critical for accurate biological interpretation and biomarker discovery.
• LNN is versatile, robust to noise, and applicable across various metabolomics platforms.