MS Identification Dr. Juan Antonio VIZCAINO PRIDE Group coordinator PRIDE team, Proteomics Services Group PANDA group European Bioinformatics Institute Hinxton, Cambridge United Kingdom EBI is an Outstation of the European Molecular Biology Laboratory. Overview … • Search engines: peptide identification • Protein inference • De novo and spectral searches • Choosing the right protein sequence DB • You need to learn many things… Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 It should not be a black box… From: Lilley et al., Proteomics, 2011 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS proteomics: Shot-gun/bottom-up approaches peptides MS/MS analysis 100 sequence database % 0 proteins 100 300 500 700 900 1100 1300 1500 1700 1900 fragmentation 100 MS analysis % 0 300 400 Juan A. Vizcaíno juan@ebi.ac.uk 500 600 700 800 900 1000 1100 m/z EBI Bulgaria Roadshow Rotterdam, 12 June 2012 2100 m/z P R O T O C O L PMF IDENTIFICATION Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Peptide Mass Fingerprinting (MS) MS analysis 100 Peptide Mass Fingerprinting (PMF) % MW 0 300 400 500 600 700 800 900 1000 1100 m/z - Each peak in the spectrum represents a peptide (or mixture of peptides) - Information about the Mass and Charge Not very used at present except for Gel Based approaches (in this case the Molecular Weight of the protein is known) Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Peptide Mass Fingerprinting (MS) in the web Aldente (Phenyx): http://www.expasy.org/tools/aldente/ ASCQ_ME: https://www.genopole-lille.fr/logiciel/ascq_me/ Bupid: http://zlab.bu.edu/Amemee/ Mascot: http://www.matrixscience.com/search_form_select.html MassSearch: http://www.cbrg.ethz.ch/services/MassSearch MS-Fit (Protein Prospector): http://prospector.ucsf.edu/prospector/mshome.htm PepMAPPER: http://www.nwsr.manchester.ac.uk/mapper/ Profound (Prowl): http://prowl.rockefeller.edu/prowl-cgi/profound.exe XProteo: http://xproteo.com:2698/ Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS/MS IDENTIFICATION Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS/MS MS analysis 100 Peptide Mass Fingerprinting (PMF) % 0 300 400 500 600 700 800 900 1000 m/z 1100 Fragmentation Peptide sequence information (on top of Mass and Charge) Juan A. Vizcaíno juan@ebi.ac.uk 100 MS/MS analysis % 0 100 300 500 700 900 1100 EBI Bulgaria Roadshow Rotterdam, 12 June 2012 1300 1500 1700 1900 2100 m/z Three types of MS/MS identification Protein database based comparison database theoretical spectrum sequence compare experimental spectrum Sequential comparison: de novo approaches database compare sequence de novo sequence experimental spectrum Spectral comparison Spectral library experimental spectrum compare experimental spectrum Modified From: Eidhammer, Flikka, Martens, Mikalsen – Wiley 2007 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS proteomics: peptide IDs and protein IDs 100 100 100 % 100 % 100 0 % 100 0 300 500 % 700 100 900 100 300 500 % 700 100 900 100 300 500 % 0 0 100 0 1100 1300 1500 1700 1900 2100 700 100 900 300 500 % 100 300 500 % 700 100 900 100 300 500 % 0 100 0 m/z 1100 1300 1500 1700 1900 2100 700 100 900 0 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 700 100 900 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 % 700 100 900 100 300 500 % 700 900 1100 1300 1500 1700 1900 2100 100 300 500 % 700 900 0 0 100 0 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 0 m/z 1100 1300 1500 1700 1900 2100 m/z MS/MS spectra proteins Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS proteomics: peptide IDs and protein IDs 100 100 100 % 100 % 100 0 % 100 0 300 500 % 700 100 900 100 300 500 % 700 100 900 100 300 500 % 0 0 100 0 1100 1300 1500 1700 1900 2100 700 100 900 300 500 % 100 300 500 % 700 100 900 100 300 500 % 0 100 0 m/z 1100 1300 1500 1700 1900 2100 700 100 900 0 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 700 100 900 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 % 700 100 900 100 300 500 % 700 900 1100 1300 1500 1700 1900 2100 100 300 500 % 700 900 0 0 100 0 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 0 m/z 1100 1300 1500 1700 1900 2100 m/z MS/MS spectra proteins Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS proteomics: peptide IDs and protein IDs 100 100 100 % 100 % 100 0 % 100 0 300 500 % 700 100 900 100 300 500 % 700 100 900 100 300 500 % 0 0 100 0 1100 1300 1500 1700 1900 2100 700 100 900 300 500 % 700 100 900 300 500 % 700 100 900 100 300 500 % 0 100 0 m/z 1100 1300 1500 1700 1900 2100 100 0 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 700 100 900 sequence database m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 % 700 100 900 100 300 500 % 700 900 1100 1300 1500 1700 1900 2100 100 300 500 % 700 900 0 0 100 0 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 0 UniProt IPI RefSeq MS/MS spectra peptides m/z 1100 1300 1500 1700 1900 2100 m/z Search engine TDMDNQIVVSDYAQ MDR LFDQAFGLPR AKPLMELIER DESTNVDMSLAQR DIVVQETMEDIDK NGMFFSTYDR GTAGNALMDGASQL proteins Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 SEARCH ENGINES Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Search engines UniProt IPI RefSeq sequence database Proteins TDMDNQIVVSDYAQMDR LFDQAFGLPR AKPLMELIER DESTNVDMSLAQR DIVVQETMEDIDK NGMFFSTYDR GTAGNALMDGASQL VDMSLAQR DIVVQETMEDIDK … Peptides 100 100 100 % 100 % 0 0 100 300 % 500 700 100 900 1100 1300 1500 1700 1900 2100 100 300 % 500 0 0 100 300 % 500 0 700 100 100 900 1100 300 % 500 0 100 1300 700 1500 100 900 1700 1100 300 % 500 1900 1300 700 700 100 900 1100 1300 1500 1700 1900 2100 1500 100 900 300 % 500 700 100 900 1100 1300 100 300 % 500 1900 1300 700 1500 m/z 1700 1900 2100 1500 Spectra m/z 1700 1900 2100 m/z 0 100 300 % 500 0 100 700 100 900 100 900 1100 1300 1500 1700 1900 2100 1100 300 % 500 100 1300 700 100 300 % 500 700 900 1100 1300 1500 1500 100 900 m/z 1700 1100 300 % 500 1900 1300 700 100 900 1700 1900 2100 100 100 300 500 700 900 1100 1300 Experimental Spectra Juan A. Vizcaíno juan@ebi.ac.uk m/z 1700 1100 300 % 500 1900 1300 2100 1500 m/z 1700 1500 700 900 1100 1900 1300 2100 1500 m/z 1700 1900 2100 m/z m/z 0 0 2100 1500 m/z 0 0 2100 m/z 1700 1100 100 0 0 2100 m/z 1700 1900 2100 100 300 500 700 900 1100 1300 1500 1700 m/z Sequence database matching EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Theoretical Spectra 1900 2100 m/z Search engines 800 1200 1600 2000 800 2400 Experimental Spectra 2000 Theoretical Spectra How good is the correlation? -Scores are generated by search engines -Usually the best match is kept juan@ebi.ac.uk 1600 m/z m/z Juan A. Vizcaíno 1200 EBI Bulgaria Roadshow Rotterdam, 12 June 2012 2400 Search engines Taken from Nesvizhskii, J Proteomics, 2010 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Search engines Taken from Nesvizhskii, J Proteomics, 2010 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 The most popular algorithms • MASCOT (Matrix Science) http://www.matrixscience.com • SEQUEST (Scripps, Thermo Fisher Scientific) http://fields.scripps.edu/sequest • X!Tandem (The Global Proteome Machine Organization) http://www.thegpm.org/TANDEM • OMSSA (NCBI) http://pubchem.ncbi.nlm.nih.gov/omssa/ Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Overall concept of scores and cut-offs Incorrect identifications Threshold score Correct identifications False negatives False positives Adapted from: www.proteomesoftware.com – Wiki pages Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Playing with probabilistic cut-off scores higher stringency 6% 100% 90% 5% 80% identifications 4% 70% 60% 3% 50% false positives 2% 40% 30% 20% 1% 10% 0% 0% p=0.05 Juan A. Vizcaíno juan@ebi.ac.uk p=0.01 EBI Bulgaria Roadshow Rotterdam, 12 June 2012 p=0.005 p=0.0005 SEQUEST • Very well established search engine • Can be used for MS/MS (PFF) identifications • Based on a cross-correlation score (includes experimental peak height) • Published core algorithm (patented, licensed to Thermo Fisher Scientific) • Provides preliminary (Sp) score, rank, cross-correlation score (XCorr), and score difference between the top tow ranks (deltaCn, Cn) • Thresholding is up to the user, and is commonly done per charge state • Many extensions exist to perform a more automatic validation of results CrossCorr XCorr = avg AutoCorr offset=-75 to 75 deltaCn= Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 XCorr 1 XCorr 2 XCorr 1 Search engines: Sequest It measures how good the XCorr is relative to the next best match. Juan A. Vizcaíno juan@ebi.ac.uk The XCorr is high if the direct comparison is significantly greater than the background EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Search engines: Mascot • Very well established search engine • Can do MS (PMF) and MS/MS (PFF) identifications • Based on the MOWSE score • Unpublished core algorithm (trade secret) • Predicts an a priori threshold score that identifications need to pass • From version 2.2, Mascot allows integrated decoy searches • Provides rank, score, threshold and expectation value per identification • Customizable confidence level for the threshold score Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Search engines: Mascot www.matrixscience.com Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Search engines: X!Tandem • Open source search engine • Can be used for MS/MS experiments • Based on a hyperscore, than only takes into account b and y ions. • Published core algorithm and it is freely available • Fast and able to handle PTMs in an iterative fashion • Used as an auxiliary search engine by-Score= Sum of intensities of peaks matching B-type or Y-type ions HyperScore= Juan A. Vizcaíno juan@ebi.ac.uk by-Score N ! N ! y EBI Bulgaria Roadshow Rotterdam, 12 June 2012 b Search engines: OMSSA • • • • • Open source search engine Can be used for MS/MS experiments Relies on a Poisson distribution Published core algorithm and it is freely available Provides an expectancy score, similar to the BLAST E-value • Very good performance in comparison with the others • Used as an auxiliary search engine Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 MS proteomics: peptide IDs and protein IDs 100 100 100 % 100 % 100 0 % 100 0 300 500 % 700 100 900 100 300 500 % 700 100 900 100 300 500 % 0 0 100 0 1100 1300 1500 1700 1900 2100 700 100 900 300 500 % 700 100 900 300 500 % 700 100 900 100 300 500 % 0 100 0 m/z 1100 1300 1500 1700 1900 2100 100 0 m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 700 100 900 sequence database m/z 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 % 700 100 900 100 300 500 % 700 900 1100 1300 1500 1700 1900 2100 100 300 500 % 700 900 0 0 100 0 1100 1300 1500 1700 1900 2100 m/z 1100 1300 1500 1700 1900 2100 m/z m/z 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 1100 1300 1500 1700 1900 2100 100 300 500 700 900 0 peptides m/z 1100 1300 1500 1700 1900 2100 m/z Search engine MS/MS spectra So far, we have actually identified peptides, not proteins Juan A. Vizcaíno juan@ebi.ac.uk UniProt IPI RefSeq EBI Bulgaria Roadshow Rotterdam, 12 June 2012 TDMDNQIVVSDYAQ MDR LFDQAFGLPR AKPLMELIER DESTNVDMSLAQR DIVVQETMEDIDK NGMFFSTYDR GTAGNALMDGASQL proteins MS proteomics: peptide IDs and protein IDs peptides proteins IPI00302927 IPI00025512 IPI00002478 IPI00185600 IPI00014537 IPI00298497 IPI00329236 IPI00002232 TDMDNQIVVSDYAQ MDRTW LFDQAFGLPR AKPLMELIER DESTNVDMSLAQR DIVVQETMEDIDK NGMFFSTYDR GTAGNALMDGASQL Protein Inference is complex!! Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 PROTEIN INFERENCE Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Intermezzo: Protein inference The minimal and maximal explanatory sets Minimal set Occam { peptide a proteins prot X prot Y prot Z x x b c d x x x x c d The Truth Maximal set anti-Occam Juan A. Vizcaíno juan@ebi.ac.uk { peptide a proteins prot X prot Y prot Z x x EBI Bulgaria Roadshow Rotterdam, 12 June 2012 b x x x x Intermezzo: Protein inference Slide from J. Cottrell, Matrix Science Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Protein inference A B C D Unambiguous peptide Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 OTHER APPROACHES TO PERFORM MS/MS IDENTIFICATION Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Three types of MS/MS identification Protein database based comparison database theoretical spectrum sequence compare experimental spectrum Sequential comparison: de novo approaches database compare sequence de novo sequence experimental spectrum Spectral comparison Spectral library experimental spectrum compare experimental spectrum Modified From: Eidhammer, Flikka, Martens, Mikalsen – Wiley 2007 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 De novo approaches Example of a manual de novo of an MS/MS spectrum No more database necessary to extract a sequence! Juan A. Vizcaíno juan@ebi.ac.uk Algorithms References Lutefisk Sherenga PEAKS PepNovo … Dancik 1999, Taylor 2000 Fernandez-de-Cossio 2000 Ma 2003, Zhang 2004 Frank 2005, Grossmann 2005 … EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Three types of MS/MS identification Protein database based comparison database theoretical spectrum sequence compare experimental spectrum Sequential comparison: de novo approaches database compare sequence de novo sequence experimental spectrum Spectral comparison Spectral library experimental spectrum compare experimental spectrum Modified From: Eidhammer, Flikka, Martens, Mikalsen – Wiley 2007 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Spectral searching • Concept: To compare experimental spectra to other experimental spectra. • There are many spectral libraries publicly available (for instance, from NIST) • Custom ‘search engines’ have been developed: • SpectraST (TPP) • X!Hunter (GPM) • It has been claimed that the searches have more sensitivity that with sequence database approaches Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Spectral searching (2) http://peptide.nist.gov/ Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 COMBINING DIFFERENT SEARCH APPROACHES Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Multi-stage peptide identification strategy Goal: “Squeeze” your good quality experimental spectra Taken from Nesvizhskii, J Proteomics, 2010 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 PROTEIN SEQUENCE DATABASES Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 What is needed from a protein database 1. Comprehensive (whatever is not in the DB will not be included in your results). 2. Not too redundant at the protein sequence level - Protein inference gets easier - It is not very good if the database is too big. 3. Quality of annotation 4. Stability of identifiers Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Main databases used a) UniProt Knowledgebase (UniProtKB): SWISS-PROT (manually curated)/ TrEMBL. b) NCBI non-redundant database: It compiles all protein sequences available from the following databases: ‘GenBank’ translations, the Protein Data Bank (PDB), UniProtKB/SwissProt, PIR and PRF. c) Ensembl: Genomics centric resource. Integration of the information with genomics is easy. d) IPI (International Protein Index): It has been discontinued (9/2012). Different builds for different species (Human, Mouse, Cow, Rat, Zebrafish, Dog, Arabidopsis). a) Model organisms DBs (for instance, TAIR for Arabidopsis). Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Databases for non-model organisms - If the species is not well represented in the protein databases, there is a much stronger need to search ESTs or genomic databases. -The search engine will translate the 6 possible ORFs for each nucleotide sequence. - ESTs are not suitable for PMF approaches (incomplete proteins). - The alternative is to filter comprehensive databases like UniProt by species or genus, or to use a protein DB from a close organism. Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Importance of choosing the right DB -Since each database has a different focus, the databases can vary in terms of completeness, degree of redundancy, and quality of annotations. -More inclusive bigger protein databases will take longer to search - For the bigger resources, it may also result on more false-positive identifications and reduced statistical significance (the probability of random match is higher). Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 POST-VALIDATION OF RESULTS Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Other concepts that would be nice to learn… -Concepts of peptide and protein FDR -Decoy databases - Softwares like PeptideProphet, ProteinProphet, … -Influence of PTMs in the search -Scoring of PTM positioning ….. Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Recommended reading…. Nesvizhskii, J Proteomics, 2010 and many more… Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Conclusions • Approaches to perform peptide and protein identification • Sequence database based approaches: search engines • The protein inference problem • Importance of choosing the right protein database • Many things to be learnt… Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Remember: it should not be a black box… From: Lilley et al., Proteomics, 2011 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 And still… we haven’t touched quantification at all From: Vaudel et al., Proteomics, 2010 Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012 Questions? Juan A. Vizcaíno juan@ebi.ac.uk EBI Bulgaria Roadshow Rotterdam, 12 June 2012