Mapping global exposure to POPs in human milk through multivariate fingerprinting

Area of Science:

• Environmental Science
• Toxicology
• Data Science

Background:

• Persistent organic pollutants (POPs) in human milk pose risks to infants.
• Traditional methods focus on POP concentrations, potentially missing compositional patterns.

Purpose of the Study:

• To apply Uniform Manifold Approximation and Projection (UMAP) for analyzing POP compositional patterns in human milk.
• To identify and compare POP 'fingerprints' across different regions, economic statuses, and time periods.

Main Methods:

• Utilized a dimensionality reduction workflow with UMAP for POP data exploration.
• Compared UMAP's effectiveness against traditional univariate, Hierarchical Cluster Analysis (HCA), and Principal Component Analysis (PCA) methods.

Main Results:

• UMAP identified detailed POP fingerprint variations undetectable by univariate methods.
• Distinct regional and economic POP fingerprints were revealed: DDT-dominant in lower-income countries, PCBs/legacy pollutants in higher-income regions.
• Temporal analysis showed POP fingerprint shifts post-2001, aligning with Stockholm Convention expansions.

Conclusions:

• UMAP effectively distinguishes compositional POP fingerprints across geospatial, temporal, and economic factors.
• This approach provides a novel comparative framework for understanding global POP exposure patterns in human milk.