Machine Learning Combined with Spectral Imaging for Inferring the Toxicity of Micro- and Nanoplastics

Introduction

The project aims to advance our understanding of potential risks posed by micro- and nanoplastics (MNPs) to human gastrointestinal health through a combination of quantitative, experimental, and computational approaches, leveraging powerful machine learning (ML) algorithms and versatile spectral imaging techniques.

Framework Development

Towards this goal, the project will first deliver a framework to extensively characterise MNPs using multiple spectral imaging techniques covering from micro- to nanoscale coupled with complementary instruments.

Data Integration and ML Models

The fused characterisation data will be combined with experimental in vitro bioassays to develop ML models, enabling the prediction of toxicity patterns and unveiling key drivers of MNP toxicity. Harnessing the broad literature data, a knowledge-based deep learning approach will be employed to unlock mechanistic insights into toxicological pathways.

Innovative Predictive Models

The most ambitious part of the proposal is to integrate previously acquired knowledge to develop innovative predictive models for predicting human health impacts of MNPs based on their physicochemical properties. This will be achieved through two independent pathways:

1. One built on insights from in vitro experiments
2. Another rooted in extensive literature data.

Ground-breaking Approaches

The ground-breaking approaches hold the potential to revolutionise the characterisation and risk assessment of MNPs, significantly reducing reliance on expensive in vitro and in vivo experiments.

Unique Integration of Competencies

This project offers a unique integration of approaches, competencies, and resources in environmental science, life science, analytical chemistry, machine learning, and computer vision, as well as technological developments of spectral imaging instruments.

Potential Outcomes

The outcomes could yield potential breakthroughs in numerous key applications of tremendous human, technological, and environmental importance, such as:

• Toxicological screening of drugs
• Safety assurance
• Environmental hazard monitoring

This project could open a whole new field of research in toxicology.