thank you for your response which has certainly enlightened me with respect
to your approach to data analysis. The big concern for me, especially for
those high MW proteins I mentioned in my previous post, was the
interpretation of the data where multiple, unrelated matches were returned
irrespective of tightening or loosening the search criteria.
The worry for me is not giving the "I don't know" result, more the wrong
one based on circumstantial but not overwhelming evidence. Your goal of
getting a single (or family of) answers is clearly the key to your success
and I should be satisfied with removing at least some of the work using
peptide mapping to eliminate those where the evidence is, indeed,
overwhelming and, as you say, getting on with the more promising spots.
Looking forward to having more of these fruitful discussion in San Diego.
Regards...Ken
>To: abrf@aecom.yu.edu@inet
>cc:
>From: Roland S Annan @ SB_PHARM_RD
>Date: 10-Mar-98 03:58:06 PM
>Subject: peptide mapping
>Categories:
>
>Ken Mitchelhill,
>
> Ken,
>
> You didn't say whether your 100 rat liver 2D gel spots were silver
>stained or coomassie stained. Here's what I'd do.
>
>Do the in-gel digest, and use 10% of the sample and get a MALDI map. If
>the data looked good, fine. If the data looked weak but promising, I'd
>concentrate the 25-100% of the remaining sample to 0.5 uL with a
>microcolumn and re-acquire the data. It's not very tough to get more rat
>liver.
>
>We acquire the data using time-lag-focusing (delayed extraction) and use
>two internal standards. The searches are done using .005-.01 % error
>tolerance. We use no MW limits and no species delimiter. The goal is to
>get a single answer, or family of answers (like 50 actins). We are 100%
>confident if we get a single answer where the matched peptides cover 20-30%
>of the sequence. This is much easier for big proteins.
>
>The way we do the first pass search, we should really get either the right
>answer or no answer. If we get no answer, we carefully back off of the
>criteria. It's possible the sample is a mixture, and then you will need to
>accept fewer matches from the total as a hit (but the coverage should still
>be no worse than 20-30%). Even this should probably provide only a single
>answer, or one that stands out from among a few (3-4 different proteins).
>
>If you have a good healthy data set and you find it necessary to back off
>until you start getting 10-25 different proteins with all the same number
>of peptide matches, you probably need to resort to sequencing.
>
>Your chances of finding matches are much poorer in the rat because so
>little of the genome is known. But the sad truth is that even very faint
>spots turn out to be actin, GST, BSA or heat shock proteins, and these are
>easily identified by MALDI mass mapping. This is done in less than an hour
>rather than wasting time sequencing junk proteins. We recently worked on
>5 "unknown" rat liver proteins from silver stained gels and identified
>them all using MALDI mass maps. Three of the five were very common
>proteins. For 100 rat liver spots I could only guess that your success
>rate might be 25-45%.
>
>In the end it's worth doing the MALDI because it's requires so little
>sample and so much less time compared to every other technique. You will
>eliminate many of the samples and can then concentrate your energies on the
>more promising spots.
>
>Roland Annan
>SmithKline Beecham
********************************
Ken I. Mitchelhill
The John Holt Protein Structure Laboratory
St. Vincent's Institute of Medical Research
41 Victoria Parade
Fitzroy 3065 Victoria
AUSTRALIA
Telephone: 61-3-9288 2480
Facsimile: 61-3-9416 2676
Email: k.mitchelhill@medicine.unimelb.edu.au
Laboratory: http://www.medstv.unimelb.edu.au/WWWDOCS/SVIMRdocs/JHPSL.html
ABRF: http://www.medstv.unimelb.edu.au/abrf.html
***********************************