A New Statistical Approach to Characterize Chemical-Elicited Behavioral Effects in High-Throughput Studies Using Zebrafish

Zebrafish have become an important alternative model for characterizing chemical bioactivity, partly due to the efficiency at which systematic, high-dimensional data can be generated. However, these new data present analytical challenges associated with scale and diversity. We developed a novel, rob...

Full description

Saved in:
Bibliographic Details
Published in:PloS one 2017-01, Vol.12 (1), p.e0169408
Main Authors: Zhang, Guozhu, Truong, Lisa, Tanguay, Robert L, Reif, David M
Format: Article
Language:English
Subjects:
Animals
Automation
Behavior
Biocompatibility
Bioinformatics
Biological activity
Biology and Life Sciences
Biometry - methods
Chemicals
Coefficient of variation
Danio rerio
Data processing
Embryos
Extreme values
Fertilization
Genomes
High-throughput screening
High-Throughput Screening Assays - methods
Laboratories
Light effects
Mathematical analysis
Mathematical models
Medicine and Health Sciences
Models, Statistical
Multivariate analysis
Neurotoxicity
Physical Sciences
Physiological aspects
R&D
Research & development
Research and Analysis Methods
Statistical methods
Statistical models
Studies
Toxicity
Toxicology
Variance analysis
Zebrafish
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Zebrafish have become an important alternative model for characterizing chemical bioactivity, partly due to the efficiency at which systematic, high-dimensional data can be generated. However, these new data present analytical challenges associated with scale and diversity. We developed a novel, robust statistical approach to characterize chemical-elicited effects in behavioral data from high-throughput screening (HTS) of all 1,060 Toxicity Forecaster (ToxCast™) chemicals across 5 concentrations at 120 hours post-fertilization (hpf). Taking advantage of the immense scale of data for a global view, we show that this new approach reduces bias introduced by extreme values yet allows for diverse response patterns that confound the application of traditional statistics. We have also shown that, as a summary measure of response for local tests of chemical-associated behavioral effects, it achieves a significant reduction in coefficient of variation compared to many traditional statistical modeling methods. This effective increase in signal-to-noise ratio augments statistical power and is observed across experimental periods (light/dark conditions) that display varied distributional response patterns. Finally, we integrated results with data from concomitant developmental endpoint measurements to show that appropriate statistical handling of HTS behavioral data can add important biological context that informs mechanistic hypotheses.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0169408