An Exploration of Machine Learning Methods in Human Biomonitoring

Area of Science:

• Environmental Health Sciences
• Computational Biology
• Toxicology

Background:

• Artificial intelligence (AI) and machine learning (ML) offer advanced capabilities for managing and analyzing large datasets.
• Human biomonitoring collects crucial data on chemical exposure and health outcomes.
• Integrating AI/ML into biomonitoring can enhance data interpretation and predictive power.

Purpose of the Study:

• To explore the implementation of AI/ML methods in human biomonitoring.
• To review current practices, applications, and researcher perceptions of AI in biomonitoring.
• To identify barriers hindering the adoption of AI in the field.

Main Methods:

• A mixed-methods approach combining a scoping literature review and an international survey of biomonitoring programs.
• The literature review identified and categorized 286 studies applying ML to human biomonitoring data.
• An online survey gathered data on AI implementation, perspectives, and barriers from 30 biomonitoring programs across 15 countries.

Main Results:

• The review identified 82 ML methods, with supervised approaches being most common, predominantly used for predicting health outcomes from chemical exposure.
• Approximately 27% of surveyed biomonitoring programs reported using AI-related methods.
• A significant barrier (80%) to AI adoption was identified as a lack of technical expertise.

Conclusions:

• Machine learning shows significant promise for advancing the understanding of chemical exposure in human populations.
• Continued growth in AI applications within human biomonitoring is anticipated.
• Addressing the technical expertise gap is crucial for broader AI adoption in biomonitoring.