Protein Sequence Databases, Peptides to Proteins, and Statistical Significance

1. Protein Sequence Databases, Peptides to Proteins, and Statistical Significance Nathan Edwards Department of Biochemistry and Mol. & Cell. Biology Georgetown University Medical Center

2. Protein Sequence Databases • Link between mass spectra and proteins • A protein’s amino-acid sequence provides a basis for interpreting • Enzymatic digestion • Separation protocols • Fragmentation • Peptide ion masses • We must interpret database information as carefully as mass spectra.

3. More than sequence… Protein sequence databases provide much more than sequence: • Names • Descriptions • Facts • Predictions • Links to other information sources Protein databases provide a link to the current state of our understanding about a protein.

4. Much more than sequence Names • Accession, Name, Description Biological Source • Organism, Source, Taxonomy Literature Function • Biological process, molecular function, cellular component • Known and predicted Features • Polymorphism, Isoforms, PTMs, Domains Derived Data • Molecular weight, pI

5. Database types

6. SwissProt • From ExPASy • Expert Protein Analysis System • Swiss Institute of Bioinformatics • ~ 515,000 protein sequence “entries” • ~ 12,000 species represented • ~ 20,000 Human proteins • Highly curated • Minimal redundancy • Part of UniProt Consortium

7. TrEMBL • Translated EMBL nucleotide sequences • European Molecular Biology Laboratory • European Bioinformatics Institute (EBI) • Computer annotated • Only sequences absent from SwissProt • ~ 10.5 M protein sequence “entries” • ~ 230,000 species • ~ 75,000 Human proteins • Part of UniProt Consortium

8. UniProt • Universal Protein Resource • Combination of sequences from • Swiss-Prot • TrEMBL • Mixture of highly curated/reviewed (SwissProt) and computer annotation (TrEMBL) • “Similar sequence” clusters are available • 50%, 90%, 100% sequence similarity

9. RefSeq • Reference Sequence • From NCBI (National Center for Biotechnology Information), NLM, NIH • Integrated genomic, transcript, and protein sequences. • Varying levels of curation • Reviewed, Validated, …, Predicted, … • ~ 9.7 M protein sequence “entries” • ~ 209,000 reviewed, ~ 90,000 validated • ~ 39,000 Human proteins

10. RefSeq • Particular focus on major research organisms • Tightly integrated with genome projects. • Curated entries: NP accessions • Predicted entries: XP accessions • Others: YP, ZP, AP

11. IPI • International Protein Index • From EBI • For a specific species, combines • UniProt, RefSeq, Ensembl • Species specific databases: HInv-DB, VEGA, TAIR • ~ 87,000 (from ~ 307,000 ) human protein sequence entries • Human, mouse, rat, zebra fish, arabidopsis, chicken, cow • Slated for closure November 2010, but still going…

12. MSDB • From the Imperial College (London) • Combines • PIR, TrEMBL, GenBank, SwissProt • Distributed with Mascot • …so well integrated with Mascot • ~ 3.2M protein sequence entries • “Similar sequences” suppressed • 100% sequence similarity • Not updated since September 2006 (obsolete)

13. NCBI’s nr • “non-redundant” • Contains • GenBank CDS translations • RefSeq Proteins • Protein Data Bank (PDB) • SwissProt, TrEMBL, PIR • Others • “Similar sequences” suppressed • 100% sequence similarity • ~ 10.5 M protein sequence “entries”

14. Human Sequences • Number of Human genes is believed to be between 20,000 and 25,000

15. DNA to Protein Sequence Derived from http://online.itp.ucsb.edu/online/infobio01/burge

16. Shows many sources of protein sequence evidence in a unified display UCSC Genome Browser

17. Accessions • Permanent labels • Short, machine readable • Enable precise communication • Typos render them unusable! • Each database uses a different format • Swiss-Prot: P17947 • Ensembl: ENSG00000066336 • PIR: S60367; S60367 • GO: GO:0003700;

18. Names / IDs • Compact mnemonic labels • Not guaranteed permanent • Require careful curation • Conceptual objects • ALBU_HUMAN • Serum Albumin • RT30_HUMAN • Mitochondrial 28S ribosomal protein S30 • CP3A7_HUMAN • Cytochrome P450 3A7

19. Description / Name • Free text description • Human readable • Space limited • Hard for computers to interpret! • No standard nomenclature or format • Often abused…. • COX7R_HUMAN • Cytochrome c oxidase subunit VIIa-related protein, mitochondrial [Precursor]

20. FASTA Format • > • Accession number • No uniform format • Multiple accessions separated by | • One line of description • Usually pretty cryptic • Organism of sequence? • No uniform format • Official latin name not necessarily used • Amino-acid sequence in single-letter code • Usually spread over multiple lines.

21. FASTA Format

22. Organism / Species / Taxonomy • The protein’s organism… • …or the source of the biological sample • The most reliable sequence annotation available • Useful only to the extent that it is correct • NCBI’s taxonomy is widely used • Provides a standard of sorts; Heirachical • Other databases don’t necessarily keep up • Organism specific sequence databases starting to become available.

23. Buffalo rat Gunn rats Norway rat Rattus PC12 clone IS Rattus norvegicus Rattus norvegicus8 Rattus norwegicus Rattus rattiscus Rattus sp. Rattus sp. strain Wistar Sprague-Dawley rat Wistar rats brown rat laboratory rat rat rats zitter rats Organism / Species / Taxonomy

24. Controlled Vocabulary • Middle ground between computers and people • Provides precision for concepts • Searching, sorting, browsing • Concept relationships • Vocabulary / Ontology must be established • Human curation • Link between concept and object: • Manually curated • Automatic / Predicted

25. Gene Ontology • Hierarchical • Molecular function • Biological process • Cellular component • Describes the vocabulary only! • Protein families provide GO association • Not necessarily any appropriate GO category. • Not necessarily in all three hierarchies. • Sometimes general categories are used because none of the specific categories are correct.

26. Gene Ontology

27. Protein Families • Similar sequence implies similar function • Similar structure implies similar function • Common domains imply similar function • Bootstrap up from small sets of proteins/domains with well understood characteristics • Usually a hybrid manual / automatic approach

28. Protein Families

29. Protein Families

30. Sequence Variants • Protein sequence can vary due to • Polymorphism • Alternative splicing • Post-translational modification • Sequence databases typically do not capture all versions of a protein’s sequence

31. Swiss-Prot Variant Annotations

32. Swiss-Prot Variant Annotations

33. Omnibus Database Redundancy Elimination • Source databases often contain the same sequences with different descriptions • Omnibus databases keep one copy of the sequence, and • An arbitrary description, or • All descriptions, or • Particular description, based on source preference • Good definitions can be lost, including taxonomy

34. Description Elimination • gi|12053249|emb|CAB66806.1| hypothetical protein [Homo sapiens] • gi|46255828|gb|AAH68998.1| COMMD4 protein [Homo sapiens] • gi|42632621|gb|AAS22242.1| COMMD4 [Homo sapiens] • gi|21361661|ref|NP_060298.2| COMM domain containing 4 [Homo sapiens] • gi|51316094|sp|Q9H0A8|COM4_HUMAN COMM domain containing protein 4 • gi|49065330|emb|CAG38483.1| COMMD4 [Homo sapiens]

35. Peptides to Proteins Nesvizhskii et al., Anal. Chem. 2003

36. Peptides to Proteins

37. 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

38. Indistinguishable Protein Sequences Nesvizhskii, Aebersold, Mol Cell Proteomics, 2005

39. Indistinguishable Protein Sequences Nesvizhskii, Aebersold, Mol Cell Proteomics, 2005

40. Protein Families Nesvizhskii, Aebersold, Mol Cell Proteomics, 2005

41. Protein Grouping Scenarios • Parsimony • Minimum # of proteins • Weighted • Choose proteinswith the most confident peptides(ProteinProphet) • Show all • Mark repeated peptides • Often no (ideal) resolution is possible! Nesvizhskii, Aebersold, Mol Cell Proteomics, 2005

42. High Quality Peptide Identification: E-value < 10-8

43. Moderate quality peptide identification: E-value < 10-3

44. Peptide Identification • Peptide fragmentation by CID is poorly understood • MS/MS spectra represent incomplete information about amino-acid sequence • I/L, K/Q, GG/N, … • Correct identifications don’t come with a certificate!

45. Peptide Identification • High-throughput workflows demand we analyze all spectra, all the time. • Spectra may not contain enough information to be interpreted correctly • …bad static on a cell phone • Peptides may not match our assumptions • …its all Greek to me • “Don’t know”is an acceptable answer!

46. What scores do “wrong” peptides get? • Generate random peptide sequences • Real looking fragment masses • Empirical distribution • Require similar precursor mass • Arbitrary score function can model anything we like!

47. Random Peptide Scores Fenyo & Beavis, Anal. Chem., 2003

48. Random Peptide Scores Fenyo & Beavis, Anal. Chem., 2003

49. Random Peptide Scores • Truly random peptides don’t look much like real peptides • Just use peptides from the sequence database! • Assumptions: • IID sampling of “score” values per spectra • Caveats: • Correct peptide (non-random) may be included • Peptides are not independent

50. Extrapolating from the Empirical Distribution • Often, the empirical shape is consistent with a theoretical model Fenyo & Beavis, Anal. Chem., 2003 Geer et al., J. Proteome Research, 2004