Stem Cell Research Paper Recalled

MattSparkes writes "One of the best-known stem cell papers describes adult cells that seemed to hold the same promise as embryonic stem cells. Now some of the data contained within the paper is being questioned, after staff at a consumer science magazine noticed errors. It shows how even peer-reviewed papers can sometimes 'slip through the net' and get to publication with inaccurate data."

Sexy science, bad science

[Ra Zen]

Stem cells are a hot topic, so people are willing to publish sloppy research or even fake date (remember Dr Hwang Woo-suk) just to get published fast and first. The same turns out to be true with other hot topics like flu research, where claims that the 1918 flu pandemic was of avian origin are severely overblown (http://www.nature.com/nature/journal/v440/n7088/f ull/nature04824.html; a free version is here: http://www.amherst.edu/~mhood/pubs.htm [amherst.edu]). Most scientifc research is solid, and most review processes work, but publications like Science and Nature often publish articles based on how sexy they are and can curtail the review process if they think the story will be a hit. Most other journals keep things steady and have very solid review proceedures.

Of course

[benhocking]

Anyone who's read a significant number of journal articles has spotted some huge errors that somehow got published. I know of one paper (not naming any names!) where in explaining how a calculation was done it had the line: 18-7=9. Clearly (from context) the line meant to say 17-8=9, but I found it humorous that such a fundamental error got past both the original authors proof-reading and the peer-review process. These things go back and forth a couple times, usually.

Peer review isn't a perfect process. It just helps reduce the noise-to-signal ratio.

Re:Data says whatever you want it to

[i kan reed]

Holy crap, I never thought of it that way, but slashdot really does support a kind of peer review, and all the comments people make about it supporting "groupthink" and such can easily be extrapolated to scientific review.

I'm not sure if that's a defence of slashdot's moderation/threading system, or if it's an attack on science as it stands today.

This is unexpected?

[Rakishi]

I'm a Stats major and professors enjoy once in a while talking about the bad stats they've seen in published papers. One such paper, in a journal that was a described as "if it publishes your paper you're nearly guaranteed tenure in the field," used statistical methods that were inherently flawed (it downright failed on simple examples).

Another one published in a prestigious journal and with a few million in government backing found 100+ genes that were significantly linked to cancer. The statistics was the type that anyone who has taken even a couple courses could find flaws in. So someone redid the analysis and found ~8 such genes at best and possibly fewer. Due to the profile of this one the proper analysis is being done as a follow-up with the original researchers help (otherwise the flaws would have been much harder to identify).

So yeah, published papers can and do have flaws but they usually they get caught after a while, the point of publishing in some ways. At the same time more researchers should release their data so it can be verified more accurately (this has its own problems as if too many people run too many methods on the same data there will be spurious results of one sort or another).