Dear TPP users,

I seek to construct a SpectraST spectral library to interrogate SWATH 
datasets.

To construct such a library, I search high-accuracy TripleTOF IDA datasets 
using TANDEM, with 50 ppm as the +-
parent errors and 100 ppm as the fragment error.  Searching the decoy 
database using the "scoring, include reverse"
TANDEM parameter, or decoyFastaGenerator-made decoys produces a very small 
number of IDs relative to the target
sequences.

For instance, with the +decoy search, I obtain 9424 assigned / 3650 unique 
/ 65 decoy matches; with the -decoy
search, I obtain 9390 assigned / 3637 unique.Here, 'maximum valid 
expectation value' set to 1e-2 for both 'output'
and 'refine' TANDEM parameters. Broadening the parent and fragment windows 
(to 200 ppm, for instance) does
not significantly increase the number of hits against the decoy dataset.

It is my understanding that the peptideProphet and interProphet classifiers 
depend on well-resolved and populous
distributions of true and false-positive IDs in order to build a useful 
discriminant. It seems that these methods would
break down given significantly fewer data for FP or TP hits.

I believe that SpectraST expects iProphet spectral probabilities / 
expectation values in order to aggregate
experimental spectra. Can you suggest a method to accurately estimate these 
probabilities given few decoy hist, instead of simply
accepting all spectral identifications?

Thank you!
Vadim

-- 
You received this message because you are subscribed to the Google Groups 
"spctools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/spctools-discuss.
For more options, visit https://groups.google.com/d/optout.

Reply via email to