PeakErazor – Calibrating MALDI Mass Spectra
Peter Højrup and Karin Hjernø
PeakErazor.exe
Department of Biochemistry and Molecular Biology, University of Southern Denmark, Odense M, Denmark.
Solution
To address these problems we have created a small, user-friendly Windowsbased program, PeakErazor, using a simple concept: By comparing all mass
values from a tryptic digest against a list of known contaminants, contaminating
peaks can easily be identified and removed. At the same time you can perform a
multipoint calibration leading to improved precision. These features leaves you
with a much larger precision in the actual PMS search. Furthermore, the program
helps you detect new contaminants either in a given dataset or after general use.
A novel development of the program enables the calibration using the mass
defect which results in the calibration of mass spectra that we hitherto was not
able to calibrate. In general the final precision is in the 30-50 ppm region.
2 - General function of PeakErazor
A
80
The average mass defect is 494 ppm.
B
1410.5
Intensity
1800.7
2299.0
1000
P
1607.7
1426.5
2244.0
772.43
2225.0
2232.9
1019.5
1219.6 1476.6
1832.7
2085.8
853.43
1147.6
1490.51707.7
1940.8
804.41
2856.2
2519.0 2760.1
1106.5
1112.6
2959.2
2871.2
2839.2
2390.9
679.55
2976.2
0
1000
3500
m/z
C:\Documents and Settings\Peter\My Documents\WORD\Teaching\BM131\Day4\spec2.massml (12:57 01/30/05)
Description: SSI 39.5
A. A MALDI mass spectrum is annotated, the mass list is copied to the clipboard.
B. The mass list is pasted into PeakErazor. The program compares it to the
currently loaded ‘Erazor list’ and calculates the difference to all ‘contaminants’
that fall within the given precision.
F
H
Y
S
M
Contaminants:
Blue: trypsin
Red: keratin
QT
W
NG
E
C. In the separate graph window the deviations are plotted against the mass. A
number of spots are seen around the x-axis, indicating a likely calibration line. The
three outliers are unchecked prior to calibrating on the five remaining values.
D. After calibration the spots are within 13 ppm (which gives a norm for the
database search). One of the outliers is directly on the +1 Da grey line and is thus a
candidate for a wrongly assigned isotope. These outlying peaks are unchecked
from the list before all unchecked mass values are copied to the clipboard for PMS.
The final protein (Calreticulin) was identified with 10 peptides (± 25 ppm). After a
two-point calibration, the same protein was identified with only 9 peptides (± 30
ppm) due to a ~10 ppm offset and a slight slope relative to the multipoint calibration.
A. The initial mass spectrum of an
HPLC separated peptide shows a
peptide peak at m/z 1815. Not being
micropurified, the spectrum shows a
number of adduct ions. The precision is
quite poor at 602 ppm.
3000
C
B. Decreasing the signal to noise ratio
to 2.5 labels a large number of peaks,
most of which are peptide related, but
of very low intensity. This peptide mass
list is copied into PeakErazor.
Calibrate on these peaks
1 tryptic, 5 keratin
An initial attempt to calibrate on the
mass defect, i.e. performing a ‘straight’
multipoint linear calibration on the
deviation from the average mass defect,
yielded a reasonable mass precision in
the order of 50-80 ppm. This was better
than no calibration, but prompted an
investigation into the nature of the mass
defect.
As the most common enzyme by far is
trypsin, an enzymatic digestion of the
entire Swiss-Prot database (164.000
proteins) was done in silico, yielding
6.55 million peptides, of which 4.1
million was in the interesting region of
500..4000 Da.
A: When the mass spectrum has been annotated (in this case using
MoverZ), copy the list of monoisotopic masses to the clipboard.
C
The mass defect
The distribution of the mass defect
for all amino acid residues shown
on a scale from 0 to 1000 ppm. The
blue line indicates the average
mass defect, the green line average
not including Cys.
B: Paste the list into PeakErazor. Note that the mass values, which fit to
values in the chosen ’Erazor list’ (e.g. known contaminants) will be checked
and listed with deviation and id.
C: In the mass vs. deviation graph it is clear to see that several of the
contaminants are located on a straight line.
D: After deselecting the outlying peak (m/z 2224) the remainder is used for
a linear calibration. After the target protein has been identified, it is digested
in silico and the peptide mass values
A
reloaded into PeakErazor as <blank>.
The identified peptides can be seen as
9 hits within 12 ppm. One mass value
(m/z 1338) can be seen on the bottom
gray line (- 1 Da) and upon inspection
of the mass spectrum, it could be seen
to be a wrongly marked isotope.
CMD = actual mass value / av. mass
defect mass
Peptide mass list of hit loaded as <blank>
9 hits within 12 ppm
Wrongly marked isotope (cluster)
B
E: After switching the graph display to ’Mass defect view’, the same raw mass
list is pasted into PeakErazor. The mass defect for each peptide is now
plotted relative to the mass and can be seen to form a diffuse sloping line.
Blue: Total CMD x 100
Figure A.
E
Deviation from mass defect calibration
of tryptic peptides based on the average
mass defect of 494 ppm. Less than 1%
of tryptic peptides above mass 1050 Da.
deviate with more than 125 ppm.
Figure D.
Adjusting the average mass defect
factor in the low and high mass regions
yields a total CMD close to zero. Also
the number of peptides >125 ppm
decreases considerably.
D. A following ’Auto’ calibration
performs a linear fit to the existing data.
A check of the resulting mass values
shows that the precision is now 57
ppm.
602 ppm
1 8 1 5 .3
1921
3 .2
1 8 5 3 .2
1 3 8 5 .3
7 1 6 .1 9
2 0 6 6 .4
0
1000
3000
m /z
\\ B m b -fils rv\ P R p u b lic \ In g e rM Z \M y o g lo b in s e p t. 0 4 \p 4 9 8 9 7 ic , to p 1 5 .m a s s m l (1 3 : 1 8 0 1 / 0 4 / 0 5 )
D e s c rip t io n : M y o g lo b . a k ta 2 2 6 4 ic , t o p 1 5
s/n 2.5
B
250
1 8 1 5 .3
1921
3 .2
1 8 5 3 .2
1 8 3 7 .2
1 3 8 5 .3
1 719894.07 .3
7 1 6 .1 9
15
7 4 58.1
28 .2
1 813807.25 .2
7 329.1
43
1 3 7 1844.4
136
7 7.1
1 .3 1 925608
54
9.2
919.2
0704.2
11.1
57
2
6 .4
1
00
.3
1 10 .3
0 .2112
.2
124
256
077.3
.3
.21 14140
7 2 7 .1 68856699.1
12146
7.2
4.4.1 1 651115777
.0
2 12.2
.2
.3
1 928
620
.3.3
02
96.0
2 163.9 31 0
0
1 00 0
3 0 00
m /z
\\ B m b - fils rv\P R p u b lic \ In g e r M Z\M y o g lo b in s e p t. 0 4 \p 4 9 8 9 7 ic , to p 1 5 . m a s s m l (1 3 :1 8 0 1 /0 4 / 0 5 )
D e s c r ip t io n : M y o g lo b . a k ta 2 2 6 4 ic , to p 1 5
C
C
manual + autocalc.
57 ppm
C
delete + recalc.
Final 51 ppm
6 - Conclusion
The current program depends on a combination of the human brain to pick out
visual trends in a graph followed by a linear multipoint calibration. Using the
‘standard’ calibration option of calibrating on contaminations a 20-50%
improvement in the precision of the mass list can usually be achieved. In
addition, contaminant mass values are removed from the search list, thus
improving the search as the mass list gets smaller.
The program saves the mass search data as either removed or included in the
search, and after several hundred spectra have been analyzed, the program
enables you to identify system specific contaminants. Another functions enables
you to identify common values in a smaller dataset (i.e. contaminants specific to
a particular gel or when analyzing isoforms/alleles).
The most recent addition to the program, mass defect calibration, comes into its
own in the cases where no internal calibrants or contaminants can be located in
the data. By taking advantage of the fact that peptide mass values are not evenly
distributed, but most values fall within ±125 ppm regions it is possible to calibrate
any peptide containing mass spectrum given that a sufficient number of peptides
are present. As a side effect, most non-peptide mass values can be removed
from the list, as they will have a mass defect larger than 125 ppm.
C
F
Figure C.
Adjusting the average mass defect for
the database residue distribution (494
ppm) brings the total CMD to zero in the
range 2000-3000 Da., but the low and
high mass ranges are not improved due
to the fact that every peptide has a Lys
or Arg, both of which have a high mass
deviation. The number of peptides with
large deviations actually gets worse.
C. The mass default deviation map
show a distribution both above and
below the x-axis. This necessitates a
manual calibration, performed by
selecting ’Manual’ followed by a click at
both ends of the low mass ’line of
deviations’ as shown.
E. Deleting the mass values with a
mass defect >125 ppm (matrix and
adduct ions) by pressing ’Delete’
followed
by
another
automatic
calibration yields the final mass values
which shows a precision of 51 ppm.
D
Omitting the mass defect of Cys from
the average mass defect (448 ppm)
improves the total CMD and lowers the
number of peptides with large
deviations.
D
m/z
H:\MSdata\P H0467A human serum proteins\p50301ic, 6_C4B_P01028.m assml (14:21 01/02/05)
Description: PH0467 A, µ-o pr. 1/3 af 6
D
Figure B.
C
3000.1
2383.6
3097.0
2500.92716.6
2704.7 2955.0
3131.9
2982.8
2932.0
A
Red: – number of peptides (in %)
deviating more than 125 ppm
2195.1
2626.8
0
Green: - Peptide distribution (x 65.500)
1442.5
2289.8 2550.8
2777.9
2052.7
881.17 1196.4 1436.5
1938.7
1851.6
2042.7
1967.7
842.421052.4
1179.5
1 338 .5
1191.5
1743.5
1833.7 2084.8
1707.5
1518.6
Calibrating on the mass defect
The figure shows:
842.541045.6
1324.5
1213 .5
1307.5
1379.5
1381.5 1638.6
1475.5
1353.5
1499.6
A
250
Intensity
966.32
K
V
R
Total CMD: Σ (Round (CMD) – CMD) /
pepCount
2283.0
A major problem when analyzing
(HPLC) isolated peptides is that your
only choice for calibration is either an
external calibration or calibrating on
matrix peaks. However, in many cases
the mass defect calibration can also be
used here – it is mainly a case of
finding
enough
(low
intensity)
monoisotopic mass peaks.
B
Intensity
A
2482.9
L
Mass defect of the 20 amino acid residues in ppm:
Ala 523
Arg 648
Asn 377
Asp 234
Cys
89
Glu 330
Gln 457
Gly 376
His 430
Leu 743
Lys 741
Met 309
Phe 465
Pro 544
Ser 368
Thr 472
Trp 426
Tyr 388
Val 691
2224.8
983.34
The mass defect is the difference between the monoisotopic
and the integer mass value of a given amino acid residue,
i.e. the mass defect of Alanine with a monoisotopic mass of
71.03711 Da is 0.03711 Da or 523 ppm.
Peptides were divided into 50 Da. slots.
For each slot a corrected mass defect
(CMD) was calculated as
2211.0
50
1000 ppm
The mass defect
5 - Isolated peptides
4 - Multipoint and mass defect calibration
3 - Compensating for variations in the mass defect
Problem
Peptide mass searching/fingerprinting (PMS) is one of the most common
methods for identifying proteins in proteomics. The method relies on identifying
proteins based on the mass values of the peptides generated by enzymatic
digestions, typically tryptic digestion. Very advanced search programs are
available today, making identification of proteins by PMS using tryptic MALDITOF MS data quite straightforward.
Two criteria are essential for a successful identification: A high number of
(significant) mass values and a high precision. In the current software we have
tried to address both problems, with most emphasis on improving the precision.
In a given dataset the PMS can be improved by removing non-significant values,
i.e. remove peptide contamination. 2-D gel separated proteins are typically
contaminated by keratin and tryptic autodigest peptides, but other project specific
contaminants may be present.
Including these peaks in the peptide mass search will obviously make the search
less precise and consequently they have to be excluded from the peptide mass
list before using the list as input in a protein identification.
Calibration is usually performed as a two- or three-point calibration (either
external or internal) of the mass spectra are used and in most cases the obtained
accuracy is sufficient for identification of the protein in question. However,
performing a multipoint internal calibration will improve the precision of the final
dataset. Furthermore, the peaks used for calibration may be missing, of poor
quality or the isotope envelope may overlap with other peaks, making internal
calibration next to impossible.
Intensity
1 - Introduction
G
The mass defect calibration does not work if the spectrum is heavily
contaminated with non-peptide peaks (i.e. matrix, detergent or adduct ions).
These cases are usually immediately recognized, as the spread of the mass
defect gets larger than 125 ppm and it is not possible to perform a ‘stable’
calibration. Few non-peptide mass values are easily removed and recalibration is
just the press of a button. Furthermore, the higher the number of peptides, the
more accurate the calibration. In praxis you need ~15 mass values for a proper
calibration. In a few cases (in our experience <5%) the method fails and we have
to resort to external calibration.
Reloading the theoretical peptide mass list of the target protein enables you to
identify wrongly assigned peptides (larger than average deviation) and helps to
locate potentially modified peptides.
D
F: Performing an linear multipoint
calibration on the mass defect with
removal of ’outliers’ (>125 ppm) yields a
diffuse scattering of points centered
around the x-axis. The mass list is now
calibrated and can be used for PMS.
8 - Where to find PeakErazor
The PeakErazor program can be downloaded for free from
G: Reloading the
target protein peptide
list shows 9 hits
within 10 ppm.
http://welcome.to/gpmaw
Please go to the ’download’ section of the web site.
If you have any problems or have suggestions for improvements, please contact the
author on [email protected].