Mapping Chemical Respiratory Sensitization: How Useful Are Our Current Computational Tools?

Area of Science:

• Toxicology and immunology, focusing on chemical respiratory sensitization.

Background:

• Chemical respiratory sensitization causes occupational asthma, a significant health issue and cause of unemployment.
• Current prediction models are lacking, and mechanisms are poorly understood, posing challenges for regulation and industry.
• Regulatory bodies often use dermal sensitization data as a proxy for respiratory sensitization due to pathway overlaps.

Purpose of the Study:

• To evaluate the accuracy of predictive models for chemical respiratory sensitization.
• To compare the performance of the Toxtree (skin sensitization) and COEH (occupational asthma) models.

Main Methods:

• Utilized a dataset of known human respiratory sensitizers.
• Assessed the accuracy of the Toxtree structural alert model and the COEH model.
• Analyzed model performance, including balanced accuracy and agreement between models.

Main Results:

• Both Toxtree and the COEH model demonstrated reasonable accuracy in predicting respiratory sensitizers.
• The COEH model achieved the highest balanced accuracy at 76%.
• When both models agreed, the overall accuracy reached 87%, with distinct strengths for each model.

Conclusions:

• The COEH model shows promise for predicting chemical respiratory sensitization.
• Combined model agreement enhances prediction accuracy.
• Addressing chemical potency and data quality is crucial for future advancements in this field.