PROTEOMICSbanner

Proteomics Viewpoint Articles

Comments to the Viewpoint article

Guidelines for the next 10 years of Proteomics
Marc R. Wilkins, Ron D. Appel, Jennifer E. Van Eyk, Maxey C. M. Chung, Angelika Görg, Michael Hecker, Lukas A. Huber, Hanno Langen, Andrew J. Link, Young-Ki Paik, Scott D. Patterson, Stephen R. Pennington, Thierry Rabilloud, Richard J. Simpson, Walter Weiss, Michael J. Dunn

Published Online: 9 Jan 2006
DOI: 10.1002/pmic.200500856


Comment from:Dr. Emanuel F. Petricoin
Short affiliation:
e-mail:epetrico@gmu.edu    Date: 2006-01-19
Comment:Dear Mike,

Congrats to you and everyone on the recent guidelines paper.
Very nice - and needed!

Comment from:Dr. Marcia Goldfarb
Short affiliation:Anatek-EP, Portland, Maine
e-mail:anatekep@maine.rr.com    Date: 2006-01-25
Comment:Dear Marc et al,

Two years ago my company, Anatek-EP gave up performing 2D and now offers analysis of 2D gels as our only product. Focusing on analysis has made me squarely face the points you discuss. While it might seem that increasing the number of replicates would insure better data, this is not necessarily the case. Poor quality replicates can increase the error. I use the program Proteomweaver which has an algorithm for determining the quality of replicate gels. I use a high cut off, and if 2 of 3 gels fall in appropriate range, I will use the 2. I will not include a replicate set which fails the test. Using that group as an average gel or using each to match singly, makes for a poorer match. My suggestion is a first step should be replicate gels scanned, matched and analyzed with a quality determination program.

I would like to suggest that replicate gels may not even be necessary. If the sample is serum, with 10 samples. The major proteins should all have the same pI amd MW in the 10 gels. An algorithm on quality of match of 20 or 30 major serum proteins can tell you if you are entitled to accept data from the set. Perhaps then a repeat of the 10 to confirm. In any case quality of match should be a first step.

I am not a statistician, and it was necessary to take on a consultant. The following is from this consultation. Data from 2D gels is not normal, and there is no transformation that will make it normal. My program JMP show this clearly. Therefore, t tests are not appropriate. The consultant suggests the "studentized range statistic" which utilizes the difference between 2 spots divided by the standard deviation.

I would like to add that most people are using premade iso strips from one of the major companies. When I was using these strips I found there could be rather large differences in length between strips in the same package. I think pressure needs to be put on these companies to have stricter QC.


back to Viewpoint Articles



©2007 Wiley-VCH Verlag GmbH & Co. KGaA - Provider
http://www.wiley-vch.de - mailto: info@wiley-vch.de