RNA-Seq Cufflinks normalization methods

Question

[Sitara]

I used tudexo to analyze the samples of drosophila melanogaster. I used Tophat2 for alignment. There are differences in the results when used with different library normalization methods. So, I am not sure which library normalization is appropriate to use. And also will be thankful if I get to know based on which criteria cufflinks uses classic FPKM and cuffdiff uses geometric as default. Any help will be appreciated. Thanks

Cufflinks with default options --library-norm-method classic-fpkm (default) for Cufflinks --library-norm-method geometric (default) for cuffdiff

Cufflinks with other options --library-norm-method classic-fpkm (default) for Cufflinks because it doesn't support other library-norm-method --library-norm-method classic-fpkm for cuffdiff

The following are the top 10 genes:

gene                      log2_fold_change_with_default      log2_fold_change_with_other
128up                         -0.197068                       -0.0925489
14-3-3epsilon                 -0.416107                       -0.323126
14-3-3zeta                    -0.582101                       -0.43963
140up                         -0.442188                       -0.296549
18SrRNA-Psi:CR41602                   0                               0
18w                           1.04193                          1.01845
26-29-p                       0.508963                         0.623343
28SrRNA-Psi:CR40596                   0                               0
28SrRNA-Psi:CR40741                   0                               0
28SrRNA-Psi:CR41609          -0.560796                         -0.579063

[WouterDeCoster]

You should know that the old 'Tuxedo' pipeline of Tophat(2) and Cufflinks is no longer the "advisable" tool for RNA-seq analysis. The software is deprecated/ in low maintenance and should be replaced by HISAT2, StringTie and ballgown. See this paper: Transcript-level expression analysis of RNA-seq experiments with HISAT, StringTie and Ballgown. There are also other alternatives, including alignment with STAR and bbmap, or pseudo-alignment using salmon.

Please stop using Tophat https://t.co/Es4ohxOEyx Cole and I developed the method in *2008*. It was greatly improved in TopHat2 then HISAT & HISAT2. There is no reason to use it anymore. I have been saying this for years yet it has more citations this year than last #methodsmatter

— Lior Pachter (@lpachter) December 2, 2017

[Sitara]

Thank you very much for the reply.

Couldn't agree more here. I have stopped using TopHat2 or Cufflinks for the analysis.

This is just a query, based on already analyzed data. All I want to know is, Cufflinks has classic FPKM as the normalization method CuffDiff has geometric normalization (default)

I have also tried, Cufflinks = Classic FPKM CuffDiff = Classic FPKM

And I have observed some differences between the two outputs. Would be great if you could advice me here, as to which one can I consider.

[WouterDeCoster]

Please use ADD COMMENT or ADD REPLY to answer to previous reactions, as such this thread remains logically structured and easy to follow. I have now moved your reaction but as you can see it's not optimal. Adding an answer should only be used for providing a solution to the question asked.

[Sitara]

Sorry. I am new to Biostar and I am learning things. Will use ADD COMMENT or ADD REPLY in the future. Could you please help me with the above query. Thank you.

Answer 1

[Devon Ryan]

FPKM normalization should never be used for statistics, it's known to be unreliable. The geometric normalization option (I think it ends up working similar to DESeq2 then) is preferred in all cases, though of course one shouldn't be using tophat2 or cufflinks to begin with.

[Sitara]

This was clear enough for me to understand. Thanks a lot for the reply.

[WouterDeCoster]

If an answer was helpful you should upvote it, if the answer resolved your question you should mark it as accepted.