You are here:
Burst and Principal Components Analyses of MEA Data Separates Chemicals by Class
Citation:
Lin, B., J. Turner, C. M. MACK, A. F. JOHNSTONE, L. D. BURGOON, AND TIM SHAFER. Burst and Principal Components Analyses of MEA Data Separates Chemicals by Class. Presented at Society of Toxicology (SOT) Annual Meeting, San Francisco, CA, March 11 - 15, 2012.
Impact/Purpose:
This demonstrates that burst analysis with PCA offers a promising approach for identifying neurotoxicological modes of action for unknown chemicals. This abstract does not reflect Agency policy.
Description:
Microelectrode arrays (MEAs) detect drug and chemical induced changes in action potential "spikes" in neuronal networks and can be used to screen chemicals for neurotoxicity. Analytical "fingerprinting," using Principal Components Analysis (PCA) on spike trains recorded from primary cortical neuron cultures is being explored to improve the utility of MEA-based high content screens to classify unknown chemicals by mode of action. The current study utilized MEA data from 10 well understood chemicals (bicuculline, lindane, RDX, picrotoxin, muscimol, verapamil, fluoxetine, chlorpyrifos oxon, domoic acid, and deltamethrin) and 3 "negative" controls (dimethyl phthalate, acetaminophen, and glyphosate) to develop PCA fingerprinting approaches. Bursting parameters (burst rate and duration, interspike intervals, #spikeslburst, etc) were computed using commercial software (NeuroExplorer,Nex Technologies) and averaged to yield parameter values as a function of concentration for each chemical, and data were standardized to the vehicle control. Burst parameter data were combined with spike rate and synchrony data and PCA was performed on these standardized mean data across all concentrations. There were a total of 16 parameters included in the PCA. The first 3 principal components accounted for 50.3,22.6, and 8.0% of the data variability. To determine how well PCA separated the high dose groups, confidence ellipsoids were drawn around data from concentrations above the chemical's IC/EC50 for spike rate. The plots illustrated ellipsoid overlapping among each of the class groups, and noticeable separation ofthe GABAA antagonists from other chemicals. As expected, the negative controls grouped with baseline activity in the absence of chemical. Domoic acid grouped closely with fluoxetine, muscimol and verapamil, while deltamethrin separated from all the other chemicals. The separation of chemicals within chemical classes and the gross separation of different chemical classes indicates that the sets may be distinguishable. This demonstrates that burst analysis with PCA offers a promising approach for identifying neurotoxicological modes of action for unknown chemicals. This abstract does not reflect Agency policy.