Biology Reference
In-Depth Information
4
Notes
1. For a complete visualization of acquired spectra, MS2 files
must be located in the same directory of the SQT files and the
processing option “Include MS2 in Results” must be checked.
Processing will take longer, but all filtered spectra and corre-
sponding ion series will be available. To convert RAW into
MS2 files, use John Yates' RawXtract program available at
2. For optimum results, we recommend fixing protein FDR to
1 % and select different configurations of spectra and peptide
FDRs. It should be noted that user-defined FDRs are not the
post-processed FDRs, which are lower than the former due to
the three-tier filtering approach.
3. Decreasing spectra and peptide FDR stringency usually results
in increasing number of proteins. However, overestimation of
the user-defined spectra and peptide FDRs results in the accep-
tance of a high number of low-scoring PSMs in the dataset.
Consequently, for achieving the 1 % protein FDR a more strin-
gent cutoff needs to be adjusted, resulting in a decrease in the
number of identified proteins.
4. Alternatively, less stringent quality filters may be applied here
in order to maximize the number of identified proteins and a
more stringent configuration may be applied later in Regrouper
before quantitative analysis is carried out.
5. Grouping PSMs by charge state and number of enzymatic ter-
mini allows peptides to be processed separately. If an insuffi-
cient number of PSMs is achieved an error message will appear.
In this situation, we recommend reducing the number of
groups by unchecking one of the grouping procedures.
6. SEPro allows the use of the semi-labeled decoy approach as
previously described [
33
].
7. Briefly, given a user-specified FDR bound, the TFold
approach uses a theoretical FDR estimator [
34
] to maxi-
mize the number of identifications that satisfy both a fold-
change cutoff, which varies with the
t
-test
p
-value as a
power law, and a stringency criterion that aims to detect
low-abundance proteins. By following these steps, the
TFold capitalizes on two limitations commonly found in
competing algorithms. A fixed fold-change cutoff could
discard proteins with very low
p
-values but not satisfying
the fold-change cutoff. Secondly, low-abundance proteins
are more prone to “fooling” common statistical filters
because they tend to artificially acquire low
p
-values and
ultimately consume the statistical power of the theoretical
FDR estimator.
