CuffLinksLibrary Normalization Methods: classic-fpkm vs -N/--upper-quartile-norm

[gwilymh]

That is a good explanation. Thank you McG

[McG]

The section of the Cufflinks manual discussing library normalization methods (http://cufflinks.cbcb.umd.edu/manual...rary_norm_meth) states that Cufflinks-supported normalization methods only work on one library at a time, while Cuffdiff-supported methods work on all of the libraries being examined. So, the quartile-norm method is similar in both Cufflinks' --upper-quartile-norm and Cuffdiff's --library-norm-method quartile methods (normalizing to the upper quartile of fragments mapping to loci). However, in cufflinks with --upper-quartile-norm, the normalization looks at the upper quartile in just the one library being examined, whereas in cuffdiff with --library-norm-method quartile, the normalization looks at the upper quartile of mapped fragments across the whole set of samples. Does that make sense?

As I'm also new to the use of Cufflinks and Cuffdiff, you can take my response with a grain of salt (and someone who is more knowledge can step in to correct me if I'm wrong).

[gwilymh]

According to the Cufflinks manual, the only option for library normalization is the "classic-fpkm" method. However, under Advanced Abundance Estimation Options there is an option to include "-N/--upper-quartile-norm" which seems similar to the "quartile" library normalization option available for CuffDiff but not CuffLinks.

I am not much of a statistician, so can anyone please explain to me the difference between the --upper-quartile-norm and --library-norm-method quartile methods.