Bayesian non-parametric non-negative matrix factorization for pattern identification in environmental mixtures

TL;DR

This paper introduces BN^2MF, a Bayesian non-parametric method for identifying chemical exposure patterns in environmental health data without pre-specifying the number of patterns, providing interpretable results with uncertainty quantification.

Contribution

The paper presents a novel Bayesian non-parametric non-negative matrix factorization approach that automatically determines the number of exposure patterns and quantifies uncertainty, improving interpretability and confidence in pattern identification.

Findings

Successfully identified chemical exposure patterns without predefining their number.

Provided variational confidence intervals to quantify uncertainty in pattern estimates.

Enhanced interpretability of patterns through non-negative priors.

Abstract

Environmental health researchers may aim to identify exposure patterns that represent sources, product use, or behaviors that give rise to mixtures of potentially harmful environmental chemical exposures. We present Bayesian non-parametric non-negative matrix factorization (BN^2MF) as a novel method to identify patterns of chemical exposures when the number of patterns is not known a priori. We placed non-negative continuous priors on pattern loadings and individual scores to enhance interpretability and used a clever non-parametric sparse prior to estimate the pattern number. We further derived variational confidence intervals around estimates; this is a critical development because it quantifies the model's confidence in estimated patterns. These unique features contrast with existing pattern recognition methods employed in this field which are limited by user-specified pattern number, lack of interpretability of patterns in terms of human understanding, and lack of uncertainty quantification.