I have 4 induced and 4 control samples (technical replicates). I ran cuffdiff to find differentially expressed genes. Now I want to create Heat Map, showing differentially expressed genes across each replicate (not condition) using Log fold change values. Is there any method to extract FPKM values for each replicate and calculate Log fold change? Or would it be appropriate to use EdgeR to perform pairwise comparison (induced1 vs control1 , induced1 vs control2 ....) between each replicate to get the log fold changes for the genes?
Note: I tried treating each replicate as a sample in cuffdiff but this method reduced the power of detecting differentially expressed genes because of the low coverage for replicates.
Any progress on this? I'm running the metrics on a bunch of distances measures right now to see the best one for my data set.
This thread is 3 years old. If you are using FPKM values and want to plot them in a heatmap, then convert them to the Z scale using zFPKM package, and then use Euclidean distance. Using Euclidean distance on the un-transformed FPKM values would be inappropriate; however, you could use something like 1 minus the correlation distance.
Se here: A: Heatmap based with FPKM values