How to identify peptides

1
How to identify peptides
Gustavo de SouzaIMM, OUS
October 2013
2
Peptide or Proteins?
3
Bottom-up Proteomics
4
2DE-based approach

5
Peptide Mass Fingerprinting
MALDI (Matrix Assisted Laser Desorption
Ionization)
6
Peptide Mass Fingerprinting
Intensity
m/z
7
MS/MS
8
MS/MS
899.013
899.013
899.013
9
Fragmentation
Nomenclature for peptide sequence-ions
Collision-Induced Dissociation (CID) MHnn
N2 --gt b y
Electron Capture Dissociation (ECD) MHnn e-
--gt MHn(n-1) --gt c z
10
Fragmentation
Roepstorff-Fohlmann-Biemann-Nomenclature
11
Fragmentation
12 aa


b ions
y ions
12
MS/MS of a peptide
y8
P y13
VPTVDVSVVDLTVK
y10
y6
y9
b5
y12
y11
y5
y4
y7
b6
b3
b10
y3
b8
b4
b7
P y13
b9
y2
b11
b12
b13
13
How to Identify MS/MS
Stenn and Mann, 2004.
Peptide Sequence Tags
Autocorrelation
Probability based match
14
Submitting to Search
15
How identification happen?
Your data
Step 1 which theoretical peptides has the same
mass of the observed ion?
Step 2 From those, which one have the most
similar fragmentation pattern?
16
High mass accuracy what is it good for?
All theoretical tryptic peptide masses from
human IPI database
Example Tryptic HSP-70 peptide ELEEIVQPIISK,
mass 1396.7813 Da
LTQ-FT
LTQ-FT
QSTAR
QSTAR
LTQ
Instrument
LTQ-FT
2 ppm
1 ppm
10 ppm
20 ppm
500
Mass Accuracy
0.5 ppm
Ext.
Ext-SIM
Int.
Ext.
Ext.
Calibration
Int.
11
9
33
52
344
of tryptic peptides for m/z 1396.7813
3
17
Defining the Search Space
18
The Search Space
2 mcl
1 mcl
0 mcl
1/2/3
1/2
2/3
2/3/4
4/5
3/4/5
1
3/4
4/5/6
2
3
5/6
4
5
1/2
2/3
1
6
3
2
4/5
4
3/4
5
5/6
6
1
3
2
4
5
6
19
Importance of Search Space Size
Search tool does not identify a peptide. It only
reports the statiscally most suitable theoretical
sequence related with the experimental data. If
you increase the size of the database too much,
or the size of the search space, false-positive
rates also increase.
20
Defining FDRs
Steen and Mann, 2004
21
MOWSE
Chance that two peptides with different sequences
but approximate Mr and sharing MS/MS similarities.
More variables inserted during search ? Higher
chance to get random events ? Higher MOWSE score
threshold

• Parameters that can modify the MOWSE calculation
• Database size
• MMD (measured mass deviation)
• Number of PTMs choosen
• Data quality.

22
Example of MMD issue

• Mycoplasma sp. sample (Munich 2006)
• Database had 700 entries
• Data accuracy had 0.7ppm average
• MMD used during search 3 ppm.

23
Strategies to Visualize FDRs
Peng et al (2003). Evaluation of multidimensional
chromatography coupled with tandem mass
spectrometry (LC/LC-MS/MS) for large-scale
protein analysis the yeast proteome. J Prot Res
2, 43-50. Reversed database sequence
24
False positive identification using reversed
database
25
Typical Result
26
How to Validate the Data
Are there any Reversed hit protein with 2
peptides above MOWSE score? -No All proteins
identified with 2 peptides score higher than
plt0.05 are good -Yes Repeat mascot search with
more stringent parameters.
What about 1-hit wonders? (Proteins identified
with only 1 peptide)
27
How to Validate the Data
Basically, the idea is to play around with the
statistics to make your result more reliable.
28
Take home message

1. Data quality (mass accuracy) and a well-defined
   search space are key for reliable peptide
   identification
2. Reliable identification is an interplay between
   asking enough without asking too much (careful
   when trying to get as many IDs as I can!)

29
PTMs
Gustavo de SouzaIMM, OUS
October 2013
30
PTMs in biology
31
PTMs in biology
32
Complexity of Protein Samples in Eukaryotes
Modifications are specificto a group of amino
acids
33
What difference to expect at MS level?
Larsen MR et al, 2006.
34
Defining the Search Space
35
PTM abundance in a cell
Total peptides in a sample
Modified peptides
Number of Peptides
Abundance level
Differences from 10e2 to 10e4
36
PTM abundance in a cell
37
Stable vs. Labile PTMs
Larsen MR et al, 2006.
38
Neutral loss
Boersema PJ et al, 2009.
39
Identifying Labile PTMs
Larsen MR et al, 2006.
40
HCD fragmentation
Larsen MR et al, 2006.
41
Status of PTM coverage
Lemeer and Heck, 2009.
42
Status of PTM coverage
Derouiche A et al, 2012.
43
Take home message

• Depending on PTM, identification can be very
  easy or very hard

• Dependent on stability under fragmentation and
  abundance in the sample
• ID improvement was mostly defined by
  instrumentationimprovements (sensitivity etc)