Skip to main content

Archived Comments for: Differential expression analysis for sequence count data

Back to article

  1. The amazing lack of proper referencing in Bioinformatics and Biostatistics

    Jean-Michel Claverie, CNRS

    17 January 2011

    As an old timer, contributing to the field of computational biology since 1975 (the previous century! :-), I am amazed by the lack of proper reference to previous works by the current generation of bioinformaticians/biostatiticians. Most of them seem to believe that nothing existed before the Internet (1990), or even later. Take the example of this article: the authors apparently ignore that before the "RNA-Seq" era (a word that appeared in the context of the NGS approaches), we were sequencing "Expressed Sequences Tags" (ESTs) exactly for the same purpose, and with similar results. Of course, theoreticians developped the proper methods to analyze these data, methods that still apply today.
    It is thus amazing that NONE of those widely used approaches are cited in this "modern work".
    And this is not by lack of visibility. For instance, the approach that I developed with my colleague Stephane Audic in 1997 (The significance of digital gene expression profiles. Genome Res 1997 7:986-95. - PMID:9331369) has been cited 500 times and is still averaging more than 40 citations per year. But this is just one example, and many highly used methods are similarly ignored. I believe it is also the responsability of the reviewers, and of the editors, to ensure that what was considered an ethical behavior not so long ago, continues to be enforced in this booming field.

    Competing interests

    None declared (except that I am citing one of my
    own article, for the sake of demonstration).

  2. Overdispersion

    Wolfgang Huber, EMBL

    8 September 2014

    Dear Jean-Michel

    apologies for the lack of citation, all I can say is that few papers manage to duly cite all relevant prior work, and that at the time of writing we must have assumed that in your case this was covered in one of the reviews or edgeR papers we cite.

    Not being cited also happens to myself all the time.

    Among the 30 or so journal and book citations we made in that paper, 6 were from the 1980s and 90s, and two from 1914 and 1953, respectively. And a main point of our approach was, of course, to model the overdispersion between biological replicates, as compared to Poisson- or multinomial sampling; an effect which I understand (Eq.(2) of your paper) seemed negligible to you in the era of EST sequencing and had only become urgent with RNA-Seq.

      Wolfgang Huber

    Competing interests

    I am an author of the paper commented here.