If you’re going to publish in the Journal of Computational Chemistry, it probably makes sense to have one’s computer tuned up for the task.
Or else you might wind up like a group of researchers in China who had to endure this misfortune:
The following article from the Journal of Computational Chemistry, “Prediction of Posttranslational Modification Sites from Sequences with Kernel Methods,” by Xiaobo Wang, Yongcui Wang, Yingjie Tian, Xiaojian Shao, Ling-Yun Wu, and Naiyang Deng, published online on 21 April 2010 in Wiley Online Library (wileyonlinelibrary.com), DOI: 10.1002.jcc.21526, has been retracted by agreement between the authors, the journal’s editors, and Wiley Periodicals, Inc. The retraction has been agreed because a computational error produced results that led the authors to overstate the level of performance of their computing model.
We’ve contacted the corresponding authors of the paper to find out more about the “computational error,” and will update with anything we find out.
When you put “computational error” in quotes, you seem to imply something more. I think this hurts what you guys are trying to do. You seem to be encouraging speculation of the uninformed. I understand that you may be quoting the term to emphasize that you are quoting the opacity of the retraction, but the quotes have more than one connotation.
Thanks for the feedback, always appreciated. As you note, however, we were simply quoting the phrase from the retraction notice that we found unclear.
It has also been noted that computational errors never UNDERESTIMATE the level of performance of any computing model.
That’s because the models that get underestimated, usually don’t get published.
Obviously, that is what I meant.
I don’t know the researchers or the journal. That said, this kind of work involves building a complicated statistical model with many moving parts. Just as it is possible to accidentally switch two rows of a PCR reaction when you’re doing hundreds of them, it is possible to accidentally write “x=y+(1*z^34+b)” instead of “x=y+(1*z^34-b)” on line 79 of a 450 line set of calculations. The problem here is not an insufficiently spiffy computer.