Natural Variation in Fish Transcriptomes: Comparative Analysis of the Zebrafish (Danio rerio)

Fathead minnow and zebrafish are among the most intensively studied fish species in environmental toxicogenomics. To aid the assessment and interpretation of subtle transcriptomic effects from treatment conditions of interest, there needs to be a better characterization and understanding of the natural variation in gene expression among fish individuals within populations. Little effort, however, has been made in this area. Leveraging the transcriptomics data from a number of our toxicogenomics studies conducted over the years, we conducted a meta-analysis of nearly 600 microarrays generated from the ovary tissue of untreated, reproductively mature fathead minnow and zebrafish samples. As expected, there was considerable batch-to-batch transcriptomic variation; this “batch-effect” appeared to impact the fish transcriptomes randomly. The overall level of variation within-batch was quite low in fish ovary tissue, making it a suitable system for studying chemical stressors with subtle biological effects. The within-batch variation, however, differed considerably among individual genes and molecular pathways. This difference in variability is probably both technical and biological, thus suggesting a need to take into account both the expression levels and variance in evaluating and interpreting the transcriptional impact on genes and pathways by experimental conditions. There was significant conservation of both the genomes and transcriptomes between fathead minnow and zebrafish. The conservation to such a degree would enable not only a comparative biology approach in studying the mechanisms of action underlying environmental stressors, but also effective sharing of a large amount of existing public transcriptomics data for future development of toxicogenomics applications. RNA from treated and control samples were directly hybridized on individual arrays. This study is a reanalysis of ovary Samples from GSE38070. The data processing is different and consisted of: Microarray image processing was performed using the latest Agilent's Feature Extraction software available at the time following the manufacturer?s protocol and recommendations. Data processing was conducted using R package Limma in the order of backgroundCorrect("normexp"), normalizeWithinArrays("loess"), and normalizeBetweenArrays("quantile"). Various types of control probes were filtered out. Each expression value represents the average of its duplicated probes. Only intensity data from control channel were utilized in the study after channel splitting. The supplementary files attached contain log2 ratios(M) and log2 intensities (A) after backgroundCorrect("normexp"), normalizeWithinArrays("loess"), and normalizeBetweenArrays("quantile") in Limma; various types of control probes were filtered