TRIumph in nanotoxicology: simplifying transcriptomics into a single predictive variable.

Area of Science:

• Toxicogenomics
• Computational Biology
• Nanotoxicology

Background:

• Transcriptomic data analysis presents challenges due to high dimensionality.
• Existing methods require substantial computational resources.
• Need for New Approach Methodologies (NAMs) to reduce animal testing.

Purpose of the Study:

• Introduce a novel transcriptomic response index (TRI) to simplify transcriptomic data.
• Link TRI with physicochemical properties of inhaled multiwalled carbon nanotubes (MWCNTs).
• Develop a predictive model for gene expression changes based on MWCNT properties.

Main Methods:

• Developed a transcriptomic response index (TRI) to compress transcriptomic space into a single variable.
• Utilized quantitative structure-activity relationship (QSAR) and nano-QSAR models.
• Trained models on fold changes of thousands of differentially expressed genes (DEGs).

Main Results:

• TRI successfully compressed 5167 DEGs into a single variable, explaining 99.9% of the transcriptomic space.
• Nano-QSAR model linking TRI and MWCNT properties achieved high statistical significance (R²=0.83, Q_CV²=0.8, Q²=0.78).
• Demonstrated the ability to predict gene expression changes using a single variable.

Conclusions:

• The TRI offers a powerful approach to manage transcriptomic data complexity.
• This methodology supports NAMs by reducing animal testing and computational load.
• Developed the ChemBioML Platform, a user-friendly tool for machine learning model development in regulatory science.