Nick Matzke posted Entry 3203 on June 22, 2007 01:21 PM.
Trackback URL: http://www.pandasthumb.org/cgi-bin/mt/mt-tb.fcgi/3192

A correction to the paper by Liu & Ochman, “Stepwise formation of the bacterial flagellar system,” was just published in PNAS. PT readers will recall that I and others had many problems with the methods and conclusions of this paper (see PT posts 1, 2, 3 for commentary and 4 for comprehensive links). The correction and brief comments are below.

Correction

EVOLUTION. For the article “Stepwise formation of the bacterial flagellar system,” by Renyi Liu and Howard Ochman, which appeared in issue 17, April 24, 2007, of Proc Natl Acad Sci USA (104:7116–7121; first published April 16, 2007; 10.1073/pnas.0700266104), the authors note that on page 7121, right column, line 4, the phrase “by using the Bl2seq program (with default options)” should instead read: “by using the Bl2seq program (with the ‘-F F’ option).” Additionally, in the Fig. 3 legend, line 3, the cutoff value 10-4 should instead appear as 9 x 10-4. Applying this cutoff yields a total of 70, not 24, hits of flagellar core proteins to nonflagellar proteins in the entire E. coli K-12 genome. These errors do not affect the conclusions of the article.

www.pnas.org/cgi/doi/10.1073/pnas.0705172104

Unfortunately the correction leaves out the well-known fact that removing the default filters (the “-F F” option on bl2seq) means that many homology hits with moderate e-values (like the paper’s original e=10-4 cutoff, and the even weaker revised cutoff of basically 10-3) will be spurious hits and not truly indicative of homology. How one can then say that this doesn’t effect the conclusions, when the main conclusion was that all of the core flagellar proteins are homologous (which was what the paper said, going beyond the long-known and well-established fact that many of the axial proteins are homologous), is beyond me.

Additionally, if the number of nonflagellar homologs to core flagellar proteins in the E. coli genome is now 70 instead of 24, due to the slightly revised cutoff, (1) this just goes to show how unstable the proffered homology identification technique is (and the correction didn’t deal with the database size issue at all), and (2) it is even more obscure than before to say, as the original paper did, that “the similarities among core proteins to one another are more common and, on average, stronger than to nonflagellar proteins” – which seems to have been the main reason in the first place for the contention that all 24 core flagellar proteins were related by duplication from a single ancestral gene. I guess the paper was arguing that 24 flagellar homologs out of 24 core flagellar proteins is a higher percentage than 24 (now 70) nonflagellar homologs out of 4,000 nonflagellar proteins, but how this is supposed to show that “core flagellar proteins all came from one gene” is a better idea than “core flagellar proteins were coopted from diverse sources” remains baffling.

Over the course of the original discussion of the Liu & Ochman paper the limitations of the blog format for critiquing published scientific papers have become more apparent to me. Although I still think blog critiques are important for public education and for rapidly spreading information within a community activists who already know a blogger, they are far less useful within the general scientific arena, where in the end one party has their conclusions published in PNAS and the other is just some random guy on the ‘net. So mostly I will wait for further publications in the scientific literature to sort this all out.

Commenters are responsible for the content of comments. The opinions expressed in articles, linked materials, and comments are not necessarily those of PandasThumb.org. See our full disclaimer.

Post a Comment

Use KwickXML formatting to markup your comments: <b>, <i>, <u> <s>, <quote author="...">, <url href="...">, etc. You may need to refresh before you will see your comment.




Remember personal info?

  


Comment #184205

Posted by meme on June 22, 2007 3:34 PM (e)

I guess “self-correcting nature of science” didn’t work this time. OK, let’s say “not yet” ;-)

Comment #184233

Posted by Dave Carlson on June 22, 2007 8:48 PM (e)

Nick - Perhaps this is wildly unrealistic and naive thinking on my part, but is there any way that you could publish a rebuttal or response to this paper?

Comment #184260

Posted by Nick (Matzke) on June 23, 2007 12:12 AM (e)

PNAS, unlike Science, doesn’t publish letters. A number of folks have written a letter critiquing the article and are shopping it around to see if a journal will publish it, but it is a little odd because it is a letter complaining about an article in a different journal. And there is another flagellum evolution paper in the works that may sort of serve as a rebuttal. The problem, though, is that if you can’t do a letter, then writing a full paper to point out “turning off filters gives you spurious hits” is itself not likely to be published, because that fact is so well-established, literally in textbooks, that it is not novel enough to be interesting to publish as research. But one way or the other something will come out eventually…

Trackback: Correction to Liu & Ochman paper

Posted by The Panda's Thumb on June 22, 2007 2:58 PM

A correction to the paper by Liu & Ochman, “Stepwise formation of the bacterial flagellar system,” was just published in PNAS. PT readers will recall that I and others had many problems with the methods and conclusions of this...

Trackback: Science at work (redux)

Posted by Uncommon Ground on June 23, 2007 8:28 AM

A couple of months ago Renyi Liu and Howard Ochman described a series of gene duplication and divergence events that could have been responsible for evolution of the core proteins in the bacterial flagellum (source). I pointed out at the...