How can I normalize the FPKM values from cufflinks out put?
Any suggestion would be appreciated
@hana FPKM values are already normalized. Check the cufflinks documentation
For what purposes? Normalization is a general term, so your question has no answer as written.
I would like to make the graph of distribution of assembled transcripts in one sample based on abundance and corresponding density. How can I get the abundance threshold and get the true positive and potential false positive transcript?
I'm assuming that you used cufflinks. If not, the following won't make any sense.
Firstly you'll need to use cuffmerge to merge the annotations and then re-estimate the FPKMs on the merged dataset (there's presumably some cuff* program that can do that, though I guess you could always just use cuffdiff or cufflinks again). If you didn't do that, it'd be difficult to properly compare the FPKMs between your samples (i.e., you'd end up with slightly different gene models between them). You can then plot the resulting FPKM estimates.
The only real way to determine a threshold for true/false positive is experimentally (though perhaps the graph will suggest a threshold). I should note that any such threshold will likely be experiment-specific.
Also, if you haven't done so, you should have a read through this paper.