Abstract
Purpose :
Microarrays and RNA-Seq have been used to characterize transcriptomes in models of optic nerve injury. We hypothesized that comparing transcriptomes from different models would yield a list of genes that were differentially expressed across multiple studies and that this group of genes would be critically important for the retinal response to injury.
Methods :
We defined a study as the collection of all expression data reported in the same experiment. Each study may contain multiple datasets that correspond to different time points, severity of disease and genotype. This analysis included 17 studies, comprised of 55 individual datasets. We analyzed the number of shared differentially expressed genes above a fold change (FC) of 2 between each pairwise comparison of both datasets and studies. The statistical significance of the number of shared genes between studies and between datasets was analyzed using Monte Carlo simulations (MCSs). Whole transcriptome hierarchical clustering on mouse optic nerve crush datasets was used to assess whether datasets would cluster by time point, study, tissue or some other factor.
Results :
The most commonly occurring genes, collagen isoforms, appeared in 11 of 17 studies. Other highly represented genes were Calbindin 2 (10/17) β-tubulin (9/17), A2M (9/17) and intracellular stress responders like ATF3 (9/17) and β and γ crystallins (9/17). 112 of 136 MCSs on studies and 1088 of 1485 MCSs on datasets gave a p-value < 0.05. Significance was not influenced in comparisons between different species, models and tissues. Whole transcriptome hierarchical clustering showed that datasets clustered by study rather than by time point, tissue or platform.
Conclusions :
Biological and technical variability amongst different studies prevented the detection of a core group of optic nerve injury response genes. However, even limited conservation of genes across heterogeneous studies lends confidence to the importance of these genes in injury response.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.