In this report you will analyze some RNA-seq data from Gulf Pipefish. The data come from brood pouch tissue of 12 adult males, half of whom were pregnant. The data file, pipefish_RNAseq_CPM.tsv, contains one row for each gene; the first and last two columns give some information about the gene, while the second through thirteenth columns give normalized RNA-seq data (in copies per million) for each gene in the 12 samples. These column names give the fish’s ID; those with “P” in their name are pregnant, and those with “N” in their name are not. The data have been curated somewhat, removing some unreliable genes.
You should write a report addressing the following:
Describe the data, briefly, and assess whether variation is likely to be well-explained by Gaussian noise.
How much does gene expression typically differ between pregnant and non-pregnant males, relative to inter-individual variation? Fit a robust model with separate mean expression values for pregnant and non-pregnant fish, and use inference on the priors of those means to answer this question. (Explanation: the question is about the distribution of mean differences between pregnant and non-pregnant fish, which is what the prior describes.) Note: if you don’t scale the data before giving it to Stan, you’ll probably run into severe convergence and speed issues.
Pick 10 genes to make a “pregnancy test” for male pipefish that would take data of this form (but only for the ten genes you choose), and make the test by fitting a robust logistic regression model for pregnancy using those ten genes.