Monday, April 16, 2007

Sin

Many scientific articles are forgettable and should had never been published. I am one of the many culprits. Even if I have written only a few articles in 22 years of research, there is not a single article I am proud of. There is a single article, instead, I should be extremely ashamed of:

A new criterion for automatic phase correction of high-resolution NMR spectra which does not require isolated or symmetrical lines

J. Magn. Reson., Ser. A - 1994, vol. 110, pp. 19-25

It correctly describes a method (called ZOE) that, unfortunately, doesn't work at all. This is what happened. After successfully testing my new ZOE on the only 1-H spectrum I had on my computer, I was so eager to publish the method that I performed no other test. You should know that in those years it was extremely painful to transfer a spectrum, especially from an old spectrometer like mine (I had set up a method to send it through the parallel printer port of the Gemini into the modem port of a Mac Plus). After having submitted the article, I had the opportunity to transfer more spectra and was disappointed by the failure of my ZOE method when applied to a large sample of spectra. It was not even the case of trying to improve it. It just doesn't work. I was so disappointed that I didn't even care about withdrawing the article. During the submission time I could think about it. I was publishing so little in my career and I didn't want to renounce to a single chance. I did just nothing and kept the secret for myself. The article was approved (without modifications!) and never demolished afterwards. I do hope that nobody else wasted her time trying to rewrite the algorithm and test it...
You shouldn't be surprised, because the literature abounds of such articles that claim to have found the (ultimate/definitive/astonishing...) solution for automatic phase correction and they never work. It's not dramatic because many chemists can't even notice a phase error as large as 5 degrees, at least in 1D spectroscopy. When I published iNMR 1.0 back in the beginning of 2006, a blogger had the bad idea of writing a review on it. The sad thing of the review was that contained a picture of a spectrum, and it was (to my eyes) terribly out of phase. I wrote to the author of the review: "I don't expect the average chemist to be a wizard in automatic phase correction, but I expect such a thing from a serious reviewer). Daniel Kaesmair removed the review from the web (but you can still find an older one related to iNMR 0.0).
What I want to say is that most chemist don't care if the phase is not perfect and tend to believe that whatever automatic algorithm they are using is OK. In reality we haven't found yet an algorithm that always work. I don't know if we'll succeed neither, because I often find some spectra that can't be phased at all, even manually (they contain some spikes or sharp solvent peaks whose phase is completely different from the rest of the spectrum). In March 2007 I advertised a new automatic phase correction algorithm, designed for metabolomic experiments in fluids, shipped with iNMR 2.1.4 (and slightly revised in version 2.1.8). It has been intensively tested. I have already applied it on more than 100 cases. Other people have also tested it and are using it for research. Today I don't care about the list of publication in my CV (I don't even have a CV...), I only try to sell the software. In theory, admitting that my method is the best on the market, I should keep it secret, cover my algorithm with a patent a make as much money as I can. In practice, however, the only way to be noticed is just to disclose the method and publish it on JMR or another journal. I don't know how this story may continue, neither this is the reason of this post.
All I have said so far is merely the premise to the news of the day. On Sunday my close friends Carlos and Stan have published a poster explaining yet another method, called eDispa. I had never seen such a complicated explanation before, at least in the field of phase correction of NMR spectra. I should try to study it a little more better tomorrow, but I am not motivated. I prefer testing the compiled program before wasting my time studying the equations. I suspect that eDispa, just like all the methods that came before it, will disappear into oblivion. The old DISPA, in particular, is one of the most overrated algorithms in the whole NMR field. They say the new eDISPA works, but this is exactly what I wrote about my ZOE in 1994. I bet that eDISPA is better than ZOE. During the week I will explain why I can't bet that it is the definitive automatic phase algorithm.

0 Comments:

Post a Comment

<< Home