Direct Analysis of Protein Complexes
55
3
Software for the Analysis of Tandem Mass Spectra
A key element to the identification of proteins in mixtures is the use of computer
software to sort through the large amount of tandem mass spectrometry data
acquired. To aid in this process, we have developed several computer programs
for the analysis of tandem mass spectrometry data. The first program uses concepts of mass spectral library searching to compare tandem mass spectra and can
be used for subtractive analyses (Yates et al. 1998).
Tandem mass spectra are compared using correlation analysis techniques. A
query CID spectrum is cross-correlated to all the reference spectra in a reference
library. The reference spectra can be a set of contaminant spectra or another LCI
MS/MS analysis. The cross-correlation scores are normalized against the autocorrelation of the query cm spectrum. A score is then computed using the following relationship:
h
CCmod X precurosrion
Searc score = 10.0 ----'--'------"------'-'-"ACmod precurosrlibrary
Scores near 0 and 10 represent poor and good matches, respectively. As part of
the comparison the precursor ion mlz value is used to initially compare spectra.
When a small value is used for the comparison, matches occur for spectra with
peptides of the same sequence. When the tolerance is decreased, spectra of peptides with small variations in their sequence can still be matched. The algorithm
seems to be a robust method for the comparison of peptide tandem mass spectra.
This method provides a good approach to compare spectra for removal of contaminants and to perform subtractive or comparative analysis. Shown in Fig.
4.1A is the database searching results for tandem mass spectra obtained from a
set of proteolytically digested Trypanosome brucei proteins. The protein mixture
is heavily contaminated with keratin producing good scoring hits of 17 of the top
35 peptides to keratin. To eliminate tandem mass spectra of keratin peptides
from the database search process, we have begun accumulating tandem mass
spectra of keratin peptides. All tandem mass spectra are compared against a contaminants library and those matching to a contaminant are removed from the
database searching queue. The results of the same Trypanosome brucei protein
analysis are shown in Fig. 4.1B after subtraction of the tandem mass spectra
matching to keratin peptides. All of the keratin tandem mass spectra have been
removed by the processing algorithm eliminating any matches to keratin.
This approach can also be used to compare the tandem mass spectra obtained
from two different forms of a protein to identify differences. In this manner, a
mutated or modified site can be identified. Fig. 4.2 shows a comparison between
normal human hemoglobin and a mutated form of hemoglobin. An equal aliquot
of each protein was again divided into 4 equal aliquots and then digested with the
following proteases; endoproteinase Glu-C, subtilisin, chymotrypsin, and trypsin. After digestion the pools of peptides for each protein were combined and
analyzed using LC/MS/MS. By using multiple proteases to digest each protein,
sequence coverage greater then 99 % could be achieved. The tandem mass spectra derived from the mutated protein were subtracted from those obtained for the
55
3
Software for the Analysis of Tandem Mass Spectra
A key element to the identification of proteins in mixtures is the use of computer
software to sort through the large amount of tandem mass spectrometry data
acquired. To aid in this process, we have developed several computer programs
for the analysis of tandem mass spectrometry data. The first program uses concepts of mass spectral library searching to compare tandem mass spectra and can
be used for subtractive analyses (Yates et al. 1998).
Tandem mass spectra are compared using correlation analysis techniques. A
query CID spectrum is cross-correlated to all the reference spectra in a reference
library. The reference spectra can be a set of contaminant spectra or another LCI
MS/MS analysis. The cross-correlation scores are normalized against the autocorrelation of the query cm spectrum. A score is then computed using the following relationship:
h
CCmod X precurosrion
Searc score = 10.0 ----'--'------"------'-'-"ACmod precurosrlibrary
Scores near 0 and 10 represent poor and good matches, respectively. As part of
the comparison the precursor ion mlz value is used to initially compare spectra.
When a small value is used for the comparison, matches occur for spectra with
peptides of the same sequence. When the tolerance is decreased, spectra of peptides with small variations in their sequence can still be matched. The algorithm
seems to be a robust method for the comparison of peptide tandem mass spectra.
This method provides a good approach to compare spectra for removal of contaminants and to perform subtractive or comparative analysis. Shown in Fig.
4.1A is the database searching results for tandem mass spectra obtained from a
set of proteolytically digested Trypanosome brucei proteins. The protein mixture
is heavily contaminated with keratin producing good scoring hits of 17 of the top
35 peptides to keratin. To eliminate tandem mass spectra of keratin peptides
from the database search process, we have begun accumulating tandem mass
spectra of keratin peptides. All tandem mass spectra are compared against a contaminants library and those matching to a contaminant are removed from the
database searching queue. The results of the same Trypanosome brucei protein
analysis are shown in Fig. 4.1B after subtraction of the tandem mass spectra
matching to keratin peptides. All of the keratin tandem mass spectra have been
removed by the processing algorithm eliminating any matches to keratin.
This approach can also be used to compare the tandem mass spectra obtained
from two different forms of a protein to identify differences. In this manner, a
mutated or modified site can be identified. Fig. 4.2 shows a comparison between
normal human hemoglobin and a mutated form of hemoglobin. An equal aliquot
of each protein was again divided into 4 equal aliquots and then digested with the
following proteases; endoproteinase Glu-C, subtilisin, chymotrypsin, and trypsin. After digestion the pools of peptides for each protein were combined and
analyzed using LC/MS/MS. By using multiple proteases to digest each protein,
sequence coverage greater then 99 % could be achieved. The tandem mass spectra derived from the mutated protein were subtracted from those obtained for the