by Peter Nollert
July 23, 2009 22:31
We all make mistakes, happens to me all the time and I hate it when I realize that those were avoidable mistakes. The range of errors we can make is huge, covering spelling errors to starting a wrong war. Somewhere in between there's the awkward class of scientific mistakes. Scientific publications are expected to be correct in the sense that what is reported is based on observations and conclusions drawn from such observations. In the process of turning X-ray diffraction data into a 3D structural model we apply software to help us along the way. What if the software we're using has errors? What if the software is not applied properly? I suspect that proper tool use in research endeavors forms one of the basic building blocks of our work. Regardless, mistakes can happen in many stages of this process and it's our job to double and triple check, to run controls to maintain a high standard of accuracy.
In a laudable commentary in Protein Science, Brian Matthews has taken it on himself to explain to the wider scientific community the mistakes that were made in the determination of structures of the ABC transporter MsbA and the EmrE multidrug transporter . Background: Chang et al. had retracted five papers and blamed an "in-house data reduction program" that apparently lead to reporting the "...wrong hand", rendering the resulting structures useless. Retracting five membrane protein structures is a big deal. So Brian takes his time and explains lucidly how F+ / F- interconversion leads to nonsense maps, not their mirror images. Eventually he suggests that the basis for the reported structures being wrong was due to "...a deficiency in the overall crystallographic analysis".
Mistakes happen, and of course it adds to the pain when you're shown the origin of your mistakes. But, as part of the scientific discourse, this is a highly welcome opportunity for all of us to learn how to avoid these particular mistakes and end up producing better science.
Thank you Brian!
Peter