Protein Identification by Sequence Database Search

1
Protein Identification by Sequence Database Search

• Nathan Edwards
• Department of Biochemistry and Mol. Cell.
  Biology
• Georgetown University Medical Center

2
Peptide Mass Fingerprint
Cut out 2D-GelSpot
3
Peptide Mass Fingerprint
Trypsin Digest
4
Peptide Mass Fingerprint
MS
5
Peptide Mass Fingerprint
6
Peptide Mass Fingerprint

• Trypsin digestion enzyme
• Highly specific
• Cuts after K R except if followed by P
• Protein sequence from sequence database
• In silico digest
• Mass computation
• For each protein sequence in turn
• Compare computer generated masses with observed
  spectrum

7
Protein Sequence

• Myoglobin GLSDGEWQQV LNVWGKVEAD IAGHGQEVLI
  RLFTGHPETL EKFDKFKHLK TEAEMKASED LKKHGTVVLT
  ALGGILKKKG HHEAELKPLA QSHATKHKIP IKYLEFISDA
  IIHVLHSKHP GDFGADAQGA MTKALELFRN DIAAKYKELG FQG

8
Protein Sequence

• Myoglobin GLSDGEWQQV LNVWGKVEAD IAGHGQEVLI
  RLFTGHPETL EKFDKFKHLK TEAEMKASED LKKHGTVVLT
  ALGGILKKKG HHEAELKPLA QSHATKHKIP IKYLEFISDA
  IIHVLHSKHP GDFGADAQGA MTKALELFRN DIAAKYKELG FQG

9
Amino-Acid Masses
Amino-Acid Residual MW Amino-Acid Residual MW
A Alanine 71.03712 M Methionine 131.04049
C Cysteine 103.00919 N Asparagine 114.04293
D Aspartic acid 115.02695 P Proline 97.05277
E Glutamic acid 129.04260 Q Glutamine 128.05858
F Phenylalanine 147.06842 R Arginine 156.10112
G Glycine 57.02147 S Serine 87.03203
H Histidine 137.05891 T Threonine 101.04768
I Isoleucine 113.08407 V Valine 99.06842
K Lysine 128.09497 W Tryptophan 186.07932
L Leucine 113.08407 Y Tyrosine 163.06333
10
Peptide Mass m/z

• Peptide Molecular Weight N-terminal-mass (0.00)
  Sum (AA masses) C-terminal-mass
  (18.010560)
• Observed Peptide m/z (Peptide Molecular Weight
  z Proton-mass (1.007825)) / z
• Monoisotopic mass values!

11
Peptide Masses

• 1811.90 GLSDGEWQQVLNVWGK
• 1606.85 VEADIAGHGQEVLIR
• 1271.66 LFTGHPETLEK
• 1378.83 HGTVVLTALGGILK
• 1982.05 KGHHEAELKPLAQSHATK
• 1853.95 GHHEAELKPLAQSHATK
• 1884.01 YLEFISDAIIHVLHSK
• 1502.66 HPGDFGADAQGAMTK
• 748.43 ALELFR

12
Peptide Mass Fingerprint
YLEFISDAIIHVLHSK
GHHEAELKPLAQSHATK
GLSDGEWQQVLNVWGK
HPGDFGADAQGAMTK
VEADIAGHGQEVLIR
HGTVVLTALGGILK
KGHHEAELKPLAQSHATK
ALELFR
LFTGHPETLEK
13
Sample Preparation for Tandem Mass Spectrometry
14
Single Stage MS
MS
15
Tandem Mass Spectrometry(MS/MS)
MS/MS
16
Peptide Fragmentation
Peptides consist of amino-acids arranged in a
linear backbone.
N-terminus
H-HN-CH-CO-NH-CH-CO-NH-CH-CO-OH
Ri-1
Ri
Ri1
C-terminus
AA residuei-1
AA residuei
AA residuei1
17
Peptide Fragmentation
18
Peptide Fragmentation
yn-i-1
-HN-CH-CO-NH-CH-CO-NH-
Ri1
Ri
bi1
19
Peptide Fragmentation
20
Peptide Fragmentation
Peptide S-G-F-L-E-E-D-E-L-K
MW ion ion MW
88 b1 S GFLEEDELK y9 1080
145 b2 SG FLEEDELK y8 1022
292 b3 SGF LEEDELK y7 875
405 b4 SGFL EEDELK y6 762
534 b5 SGFLE EDELK y5 633
663 b6 SGFLEE DELK y4 504
778 b7 SGFLEED ELK y3 389
907 b8 SGFLEEDE LK y2 260
1020 b9 SGFLEEDEL K y1 147
21
Peptide Fragmentation
22
Peptide Identification

• Given
• The mass of the precursor ion, and
• The MS/MS spectrum
• Output
• The amino-acid sequence of the peptide

23
Sequence Database Search
24
Sequence Database Search
25
Sequence Database Search
26
Sequence Database Search

• No need for complete ladders
• Possible to model all known peptide fragments
• Sequence permutations eliminated
• All candidates have some biological relevance
• Practical for high-throughput peptide
  identification
• Correct peptide might be missing from database!

27
Peptide Candidate Filtering

• Digestion Enzyme Trypsin
• Cuts just after K or R unless followed by a P.
• Basic residues (K R) at C-terminal attract
  ionizing charge, leading to strong y-ions
• Average peptide length about 10-15 amino-acids
• Must allow for missed cleavage sites

28
Peptide Candidate Filtering

• gtALBU_HUMAN MKWVTFISLLFLFSSAYSRGVFRRDAHKSEVAHRFKDL
  GEENFKALVLIAFAQYLQQCPFEDHVKLVNEVTEFAK

No missed cleavage sites
MK WVTFISLLFLFSSAYSR GVFR R DAHK SEVAHR FK DLGEENF
K ALVLIAFAQYLQQCPFEDHVK LVNEVTEFAK
29
Peptide Candidate Filtering

• gtALBU_HUMAN MKWVTFISLLFLFSSAYSRGVFRRDAHKSEVAHRFKDL
  GEENFKALVLIAFAQYLQQCPFEDHVKLVNEVTEFAK

One missed cleavage site
MKWVTFISLLFLFSSAYSR WVTFISLLFLFSSAYSRGVFR GVFRR RD
AHK DAHKSEVAHR SEVAHRFK FKDLGEENFK DLGEENFKALVLIAF
AQYLQQCPFEDHVK ALVLIAFAQYLQQCPFEDHVKLVNEVTEFAK
30
Peptide Candidate Filtering

• Peptide molecular weight
• Only have m/z value
• Need to determine charge state
• Ion selection tolerance
• Mass for each amino-acid symbol?
• Monoisotopic vs. Average
• Default residual mass
• Depends on sample preparation protocol
• Cysteine almost always modified

31
Peptide Molecular Weight
32
Peptide Molecular Weight
from Isotopes An IonSource.Com Tutorial
33
Peptide Molecular Weight

• Peptide sequence WVTFISLLFLFSSAYSR
• Potential phosphorylation? S,T,Y 80 Da

WVTFISLLFLFSSAYSR 2018.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2098.06
WVTFISLLFLFSSAYSR 2178.06
WVTFISLLFLFSSAYSR 2178.06

WVTFISLLFLFSSAYSR 2418.06

• 7 Molecular Weights
• 64 Peptides

34
Peptide Scoring

• Peptide fragments vary based on
• The instrument
• The peptides amino-acid sequence
• The peptides charge state
• Etc
• Search engines model peptide fragmentation to
  various degrees.
• Speed vs. sensitivity tradeoff
• y-ions b-ions occur most frequently
• The scores have no apriority scale

35
Peptide Identification

• High-throughput workflows demand we analyze all
  spectra, all the time.
• Spectra may not contain enough information to be
  interpreted correctly
• ...cell phone call drops in and out
• Spectra may contain too many irrelevant peaks
• bad static
• Peptides may not match our assumptions
• its all Greek to me
• Dont know is an acceptable answer!

36
Peptide Identification

• Rank the best peptide identifications
• Is the top ranked peptide correct?

37
Peptide Identification

• Rank the best peptide identifications
• Is the top ranked peptide correct?

38
Peptide Identification

• Rank the best peptide identifications
• Is the top ranked peptide correct?

39
Peptide Identification

• Incorrect peptide has best score
• Correct peptide is missing?
• Potential for incorrect conclusion
• What score ensures no incorrect peptides?
• Correct peptide has weak score
• Insufficient fragmentation, poor score
• Potential for weakened conclusion
• What score ensures we find all correct peptides?

40
Statistical Significance

• Cant prove particular identifications are right
  or wrong...
• ...need to know fragmentation in advance!
• A minimal standard for identification scores...
• ...better than guessing.
• p-value, E-value, statistical significance

41
Random Peptide Models

• "Generate" random peptides
• Real looking fragment masses
• No theoretical model!
• Must use empirical distribution
• Usually require they have the correct precursor
  mass
• Score function can model anything we like!

42
Random Peptide Models
Fenyo Beavis, Anal. Chem., 2003
43
Random Peptide Models
Fenyo Beavis, Anal. Chem., 2003
44
Random Peptide Models

• Truly random peptides dont look much like real
  peptides
• Just use (incorrect) peptides from the sequence
  database!
• Caveats
• Correct peptide (non-random) may be included
• Homologous incorrect peptides may be included
• (Incorrect) peptides are not independent

45
Extrapolating from the Empirical Distribution

• Often, the empirical shape is consistent with a
  theoretical model

Geer et al., J. Proteome Research, 2004
Fenyo Beavis, Anal. Chem., 2003
46
False Positive Rate Estimation

• A form of statistical significance
• Search engine independent
• Easy to implement
• Assumes a single threshold for all spectra
• Best if E-value or similar is used to compute a
  spectrum normalized score

47
False Positive Rate Estimation

• Each spectrum is a chance to be right, wrong, or
  inconclusive.
• At any given threshold, how many peptide
  identifications are wrong?
• Computed for an entire spectral dataset
• Given identification criteria
• SEQUEST Xcorr, E-value, Score, etc., plus...
• ...threshold
• Use decoy sequences
• random, reverse, cross-species
• Identifications must be incorrect!

48
Decoy Search Strategies

• Concatenated target decoy
• Competition for best hit...
• Masks good decoy scores due to spectral variation
• Separate searches
• Cleaner estimation of false hit distribution
• More conservative than concatenation
• Must ensure
• Decoy searches do not change target peptide
  scores
• Single score distribution across dataset

49
Decoy Search Strategies

• Reversed Decoys
• Captures redundancy of peptide sequences
• Susceptible to mass-shift anomalies
• Bad choice for protein-level statistics
• Shuffled Random Decoys
• Multiple independent decoys can be created.
• Better estimation of tail probabilities
• More conservative than reversed decoys

50
False Positive Rate Estimation Concatenated
Target Decoy

• Choose a threshold t.
• Count of (rank 1) target ids (Tt) with score
  t.
• Count of (rank 1) decoy ids (Dt) with score
  t.
• Compute FPR ( 2 x Dt ) / ( Tt Dt )
• Principle
• Decoy peptides equally likely as false hits at
  rank 1
• Issues
• What to do with decoy hits?
• Change in database size may affect scores

51
False Positive Rate Estimation Separate Decoy
Search

• Choose a threshold t.
• Count of (rank 1) target ids (Tt) with score
  t.
• Count of (rank 1) decoy ids (Dt) with score
  t.
• Compute FPR Dt / Tt
• Principle
• Find the distribution of false hit scores, apply
  to target
• Issues
• Can choose to merge after the fact...
• Decoy search cannot change target scores
• A few good decoy scores can inflate small FDR
  values

52
Peptide Prophet

• Re-analysis of SEQUEST results
• Spectrum dependant scores (XCorr)
• Additional features form discriminant score
• Assumes that many of the spectra are not
  correctly identified
• These identifications act like decoy hits

53
Peptide Prophet
Keller et al., Anal. Chem. 2002
Distribution of spectral scores in the results
54
Peptides to Proteins
Nesvizhskii et al., Anal. Chem. 2003
55
Peptides to Proteins
56
Peptides to Proteins

• A peptide sequence may occur in many different
  protein sequences
• Variants, paralogues, protein families
• Separation, digestion and ionization is not well
  understood
• Proteins in sequence database are extremely
  non-random, and very dependent
• No great tools for assessing statistical
  confidence of protein identifications.

57
Mascot MS/MS Ions Search
58
Mascot MS/MS Search Results
59
Mascot MS/MS Search Results
60
Mascot MS/MS Search Results
61
Mascot MS/MS Search Results
62
Mascot MS/MS Search Results
63
Mascot MS/MS Search Results
64
Mascot MS/MS Search Results
65
Sequence Database SearchTraps and Pitfalls

• Search options may eliminate the correct peptide
• Precursor mass tolerance too small
• Fragment m/z tolerance too small
• Incorrect precursor ion charge state
• Non-tryptic or semi-tryptic peptide
• Incorrect or unexpected modification
• Sequence database too conservative
• Unreliable taxonomy annotation

66
Sequence Database SearchTraps and Pitfalls

• Search options can cause infinite search times
• Variable modifications increase search times
  exponentially
• Non-tryptic search increases search time by two
  orders of magnitude
• Large sequence databases contain many irrelevant
  peptide candidates

67
Sequence Database SearchTraps and Pitfalls

• Best available peptide isnt necessarily correct!
• Score statistics (e-values) are essential!
• What is the chance a peptide could score this
  well by chance alone?
• The wrong peptide can look correct if the right
  peptide is missing!
• Need scores (or e-values) that are invariant to
  spectrum quality and peptide properties

68
Sequence Database SearchTraps and Pitfalls

• Search engines often make incorrect assumptions
  about sample prep
• Proteins with lots of identified peptides are not
  more likely to be present
• Peptide identifications do not represent
  independent observations
• All proteins are not equally interesting to report

69
Sequence Database SearchTraps and Pitfalls

• Good spectral processing can make a big
  difference
• Poorly calibrated spectra require large m/z
  tolerances
• Poorly baselined spectra make small peaks hard to
  believe
• Poorly de-isotoped spectra have extra peaks and
  misleading charge state assignments

70
Summary

• Protein identification from tandem mass spectra
  is a key proteomics technology.
• Protein identifications should be treated with
  healthy skepticism.
• Look at all the evidence!
• Spectra remain unidentified for a variety of
  reasons.

71
Further Reading

• Matrix Science (Mascot) Web Site
• www.matrixscience.com
• Seattle Proteome Center (ISB)
• www.proteomecenter.org
• Proteomic Mass Spectrometry Lab at The Scripps
  Research Institute
• fields.scripps.edu
• UCSF ProteinProspector
• prospector.ucsf.edu