by
lmh » Thu Oct 24, 2013 2:52 pm
Peter's right. But to make your job easier, make sure you prepare the samples replicated as exactly as possible, and give all manufacturers identical materials. Similarly, if you know how you want the samples run, it's a good idea to specify this very precisely, so you don't end up in the situation where you have 4 completely incomparable sets of results because they've used outrageously different methods (resolution won't mean a thing if they've all chosen different columns, sensitivity is meaningless if they all inject in wildly different ways). By all means let them run samples twice, once the way you specify, and a second time according to what they think will show off their instrument to its best, but make sure you get comparable data. You want to test the instrument, not the company's demonstration chemist! (I remember one of the managers where I work once commenting "I'm not sure if we want their instrument or not, but I wish we could get Mr XXXX to come and work for us!").
If you're doing metabolomics, do check on their software too. It may be that you already have a downstream processing pipeline that you want to use (XCMS or whatever) in which case you need to know how easily you can get their data into your pipeline; but it's good to know if they can sell you specialist metabolomic software too. If you're doing profiling-metabolomics, where you're actually looking at a (large) range of targeted peaks, then just check you're happy with their standard analytical software.
Good luck!
