Concentration-response evaluation of ToxCast compounds for multivariate activity patterns of neural network function

Area of Science:

• Neuroscience
• Toxicology
• Computational Biology

Background:

• The US Environmental Protection Agency's ToxCast program lacks comprehensive neurotoxicity assessment.
• Networks of neurons on microelectrode arrays (MEAs) offer efficient screening for neuroactivity.
• Previous studies screened single concentrations of compounds for effects on mean firing rate (MFR).

Purpose of the Study:

• Expand neuroactivity screening using concentration-response data across multiple network activity parameters.
• Characterize chemical activities relevant to neurotoxicity using multivariate MEA data.
• Identify structure-activity relationships for neurotoxic compounds.

Main Methods:

• Retested 384 compounds in concentration-response across 43 network activity parameters.
• Applied hierarchical clustering and machine learning to identify crucial activity parameters.
• Utilized k-means clustering to identify multivariate MEA response patterns.
• Analyzed MEA clusters for enrichment of chemical features (ToxPrint chemotypes).

Main Results:

• Identified 15 crucial network activity parameters for 237 active compounds.
• Neurotoxic compounds showed higher potency in MEA assays compared to other ToxCast assays.
• Discovered three distinct k-means clusters representing multivariate MEA response patterns.
• Found differential segregation of known neurotoxicants within the identified clusters.
• Revealed chemical structural features distinguishing the MEA clusters.

Conclusions:

• Multivariate MEA activity patterns efficiently screen for diverse chemical activities related to neurotoxicity.
• MEA response patterns demonstrate predictive value for chemical structural features.
• This approach enhances neurotoxicity assessment beyond current ToxCast capabilities.