Bioinformatics and machine learning to support nanomaterial grouping.

Area of Science:

• Nanomaterial toxicology and risk assessment.
• Application of bioinformatics and artificial intelligence in hazard identification.

Background:

• Nanomaterials (NMs) possess tunable physicochemical properties, leading to a vast number of variants.
• Efficient hazard and risk assessment strategies are crucial, with New Approach Methodologies (NAMs) gaining prominence.
• Grouping and read-across strategies, based on structural and physicochemical similarity, are promising for managing numerous NM variants.

Purpose of the Study:

• To demonstrate the utility of bioinformatics, specifically Machine Learning (ML), in understanding NM Modes-of-Action (MoA).
• To identify NM properties relevant to specific hazards, thereby supporting grouping strategies.
• To review existing ML models for NM toxicity prediction and omics data analysis in nanotoxicology.

Main Methods:

• Review of existing Machine Learning (ML) models for predicting nanomaterial toxicity.
• Analysis of omics datasets using ML to identify Modes-of-Action (MoA) patterns.
• Exploration of Artificial Intelligence (AI) approaches like Natural Language Processing and image analysis for information extraction.

Main Results:

• ML effectively identifies critical physicochemical properties linked to specific NM hazards.
• ML facilitates the analysis of large omics datasets, revealing MoA patterns that inform grouping hypotheses.
• Omics approaches enable a shift towards mechanistic understanding across multiple endpoints from single experiments.

Conclusions:

• Bioinformatics and ML are powerful tools for unraveling NM MoA and supporting data-driven grouping strategies.
• These approaches enhance the efficiency and reliability of nanomaterial risk assessment within NAM frameworks.
• Addressing challenges in building robust nanotoxicology models is essential for advancing AI applications in this field.