Remove any miRNA that is detected in 3 or fewer samples.
Replace any undetected value (Inf) with the average of the expression of that miRNA in the rest of the samples.
Print the first 5 rows (genes) and first 6 columns (samples) of your data.
You should end up with data that is the same as the filtered Excel data. If you get different filtered data, and cannot find out why you are getting different results, abandon your data and use the filtered Excel data.
Identify which genes may be good normalizers for this dataset. Use sum-prenormalization, followed by calculation of standard deviation. Create a scatter-plot showing each gene, with x axis the average expression across all samples and the y axis the standard deviation across all samples. Among the miRNAs that have average CT⇐30, report the top 3 miRNAs with the best standard deviation across all samples.
For each sample, find the CT0 values by averaging the CT values of RNU44, RNU48, and MammU6. Subtract the CT values of that sample with the CT0 of that sample.
The result of the previous step will give you normalized deltaCT values.
For each miRNA, find the average of all healthy deltaCT values and separately, find the average of the patient deltaCT values.
For each miRNA, subtract the average healthy deltaCT values from the average patient deltaCT values. This gives you the deltadeltaCt.
Calculate 2^-deltadeltaCT, which are the fold change values.
Replace any fold change value x that is less than 1, with its negative inverse (-1/x).
Print the names and fold changes of the top-10 most changing miRNAs (either up or down regulation, i.e., order by descending absolute fold change).
Using the deltaCT values, find the significantly different miRNAs between controls and patients.
Print the names and p-values of the top-10 most significantly different miRNAs (ordered by pvalue).
The print-out of genes requested above are limited to keep your output small. You will need to define your own p value and or fold change thresholds to be used for selecting significantly different miRNAs for the steps below. Use all differentially expressed genes for the steps below, not just the ones you printed out above.
Find which mRNAs are the predicted targets of the significant miRNAs from the CRPS study using TargetScan.
Perform enrichment of the targets using the DAVID webservice. List the top 3 most significantly enriched pathways and top 3 most significantly enriched Gene Ontology terms, along with their p-values and the number of genes from your gene list found in these pathways and terms.
Gene set enrichment with tens of genes will not get any enriched annotations and the annotations from many thousands of genes will not be meaningful. For targetscandb, you can additionally adjust the confidence score threshold to control how many genes you get. You can also adjust pvalue or fold change thresholds above to control the number of most-different micrornas and thus to limit the resulting gene set.