Two different treatments produce two distinct Th2-type responses

“

This was my first big bioinformatics project at the Malaghan Institute, and the bioinformaticians were able to provide input throughout most of this project, including on the experimental design prior to sequencing. One of AATS' initial contributions was to run simulations to work out how many mice we needed to pool in order to get sufficient biological variation (and smoothing of that variation) from three replicates per sample. I used my experience with RNASeq and Illumina sequencing to decide on what sequencing parameters were needed to make sure our sequence data would be of sufficient quality and coverage to get good results. We understood that we wouldn't be able to pick up changes in low-expression genes, but hoped that there would be enough of a sniff of differential expression in the high-expression genes that our biological questions could be answered. I also developed a Shiny app (based on graphs that AATS had made) that allowed me to manage the data, and allowed the biologists to manage the results. This application provided other authors with many of the tools they needed to quickly explore the data themselves, without having to ask me for a specific view of the data set. Most of the RNASeq results figures in the paper have been taken from this application, with only minimal modification for publication. The shiny app includes the following views of the data: * Heatmap of expression (via VSTPk) and differential expression (DESeq2 results) * Volcano plot * MA plot * Venn diagram * Scatter plot * PCA Biologists are able to select their preferred statistical model (data set), subset on a particular set of genes, and choose which treatments to display in graphs. Many of the graphs (including the venn diagram) allow the user to select a region of the graph and display the names of the selected genes. There is also an ability to download the displayed data for loading into other analysis programs (e.g. IPA). The Shiny app is supplemented by a genome browser, allowing the biologists to look at the raw data to see if/when there were any issues with the results. This lead to comments like "gene X looks like it's expressed on JBrowse, but it's not showing up as significant in the Shiny app", which were immensely helpful for correcting bugs in the scripts used to analyse the data.

”