Biological Databases for Protein Sequence Analysis

1
Biological Databasesfor Protein Sequence Analysis

• Teresa K. Attwood
• School of Biological Sciences
• University of Manchester, Oxford Road
• Manchester M13 9PT, UK
• http//www.bioinf.man.ac.uk/dbbrowser/

2
Overview

• Introduction
• Web practical, science fact fiction, some
  reality checks
• Biological databases
• sequence, family, composite, etc.
• Pattern recognition
• regular expressions, fingerprints, profiles, etc.
• Building a search protocol
• a real example

3
Introduction

• Single- three-letter amino acid codes
• G Glycine Gly P Proline Pro
• A Alanine Ala V Valine Val
• L Leucine Leu I Isoleucine Ile
• M Methionine Met C Cysteine Cys
• F Phenylalanine Phe Y Tyrosine Tyr
• W Tryptophan Trp H Histidine His
• K Lysine Lys R Arginine Arg
• Q Glutamine Gln N Asparagine Asn
• E Glutamic Acid Glu D Aspartic Acid Asp
• S Serine Ser T Threonine Thr
• Additional codes
• B Asn/Asp Z Gln/Glu X Any amino acid

4
(No Transcript)
5
Basic definitions

• Primary structure
• the linear sequence of amino acids in a protein
• Secondary structure
• regions of local regularity
• i.e., a-helices, b-strands, -sheets -turns

6
Definitions contd.

• Super-secondary structure
• the packing of secondary structure elements into
  stable units
• e.g., b-barrels, bab units, Greek keys, etc..

7
Definitions contd.

• Tertiary structure
• the overall chain fold that results from packing
  of secondary structure elements

8
Definitions contd.

• Quaternary structure
• the arrangement of separate chains within a
  protein that has more than one subunit
• e.g., haemoglobin

9
Definitions contd.

• Quinternary structure
• the arrangement of separate molecules, such as in
  protein-protein or protein-nucleic acid
  interactions

10
The practical - BioActivity

• BioActivity sequence analysis in action
• begin with a fragment of a DNA sequence
• try to find out what protein this codes for, the
  family to which it belongs, whether its
  function structure are known
• The practical is entirely Web-based
• be mindful of traffic don't waste time on slow
  links
• Most important of all
• read the instructions!
• The Web is constantly evolving....
• please report dead links (otherwise theyll stay
  dead)!

11
Importance of sequence analysis

• gt900,000 sequences available in public dbs
• millions more (including ESTs) in proprietary
  dbs
• these s will snowball with completion of more
  genomes
• so what?
• Locked up in sequences is a huge amount of
  structural, functional evolutionary info
• they're a highly valuable resource
• By contrast, the of unique protein structures
  is 2000
• a huge information deficit

12
The legacy of the genome projectsSequence-structu
re deficit
800 700 600 500 400 300 200 100
1988

2002

• Non-redundant growth of sequences during
  1988-2002 ( ) the corresponding growth in
  the number of structures ( ).

13
Challenges for bioinformatics

• Spurred on by the seq/structure deficit, the
  challenges
• rationalise the mass of sequence data
• derive more efficient means of data storage
• design more incisive reliable analysis tools
• The imperative - to convert sequence information
  into biochemical biophysical knowledge
• to decipher the structural, functional
  evolutionary clues encoded in the language of
  biological sequences

14
The Holy Grail of bioinformatics

• ...to be able to understand the words in a
  sequence sentence that form a particular protein
  structure

15
The reality of sequence analysis

• ...isn't so glamorous....but means we can
  recognise words that form characteristic
  patterns, even if we don't know the precise
  syntax to build complete protein sentences

16
Pattern recognition prediction

• In investigating the meaning of sequences, two
  distinct analytical approaches have emerged
• pattern recognition is used to detect similarity
  between sequences hence to infer related
  structures functions
• ab initio prediction is used to deduce structure,
  to infer function, directly from sequence
• These methods are quite different!
• pattern recognition methods demand that some
  characteristic has been seen before housed in a
  db
• prediction methods remove the need for template
  dbs, because deductions are made directly from
  sequence

17
Science fact fiction

• Sequence pattern recognition is easier to
  achieve, is much more reliable, than fold
  recognition
• which is 50 reliable even in expert hands
• Prediction is still not possible
• is unlikely to be so for decades to come (if
  ever)
• Structural genomics will yield representative
  structures for many (but not all) proteins in
  future
• structures of new sequences will be determined by
  modelling
• prediction will become an academic exercise
• But, to debunk a popular myth, knowing structure
  alone does not inherently tell us function

18
A reality check

• What is the function of this structure?

• What is the function of this sequence?

• What is the function of this motif?
• the fold provides a scaffold, which can be
  decorated in different ways by different
  sequences to confer different functions knowing
  the fold function allows us to rationalise how
  the structure effects its function at the
  molecular level

19
A test case for structural genomics
Structure-based assignment of the biochemical
function of hypothetical protein mj0577
(Zarembinski et al., PNAS 95 1998)
Although the structure co-crystallised with ATP,
the biochemical function of the protein is
unknown
20
The Twilight Zone

• Prediction methods dont work because we dont
  fully understand the Folding Problem
• we cant read the language sequences use to
  create their folds
• But, with sequence analysis techniques, we can
  try to find similarities between new sequences
  those in dbs
• whose structures functions we hope have been
  elucidated
• This is straightforward at high levels of
  identity, but below 50 it is difficult to
  establish relationships reliably
• Analyses can be pursued with decreasing certainty
  towards the Twilight Zone
• 20 identity, where results may look plausible
  to the eye, but are no longer statistically
  significant

21
Beyond the Twilight Zone

• To penetrate deeper into the Twilight Zone is the
  aim of most analytical methods
• whether using single sequences, motifs, complex
  weighting schemes or raw amino acid frequencies
• Each offers a different perspective, depending on
  the type of information used in the search
• none gives the right answer
• It is good practice to devise an analysis
  protocol that uses a variety of methods
• but dont expect the impossible no method is
  infallible!

22
Application areas of analysis tools

• The scale indicates identity between aligned
  sequences
• Alignment of 2 random seqs can produce 20
  identity
• less than 20 does not constitute a significant
  alignment
• around this threshold is the Twilight Zone,
  where alignments may appear plausible to the eye,
  but cant be proved by conventional methods

23
Homology analogy

• The term homology is confounded abused!
• sequences are homologous if they are related by
  divergence from a common ancestor
• analogy relates to the acquisition of common
  features from unrelated ancestors via convergent
  evolution
• e.g., b-barrels occur in soluble membrane
  proteins enzymes chymotrypsin subtilisin share
  groups of catalytic residues, with near identical
  spatial geometries, but no other similarities
• It is not a measure of similarity is not
  quantifiable
• it is an absolute statement that sequences have a
  divergent rather than a convergent relationship
• the phrases "the level of homology is high" or
  "the sequences show 50 homology", or any like
  them, are strictly meaningless!
• This is not just a semantic issue
• loose use muddies thinking about evolutionary
  relationships

24
A terminology muddle

• The same arguments apply to 3D structures
• structures may be similar, as denoted by RMS
  positional deviation between compared atomic
  positions
• but their common evolutionary origin is a
  hypothesis
• the hypothesis may be correct or mistaken, but
  their similarity is a fact, no matter how it is
  interpreted
• Similarity of sequence or structure is just that
  similarity
• Homology connotes a common evolutionary origin
• Reeck, G.R., de Haen, C., Teller, D.C.,
  Doolittle, R.F., Fitch, W.M., Dickerson, R.E.,
  Chambon, P., McLachlan, A.D., Margoliash, E.,
  Jukes, T.H. Zuckerkandl, E. (1987) Homology
  in proteins and nucleic acids a terminology
  muddle and a way out of it. Cell, 50, 667.

25
More challenges for sequence analysis

• Much of the challenge is in getting the biology
  right
• this is complicated by the problem of orthology
  vs paralogy
• Following a search, how much functional
  annotation can be legitimately inherited by a
  query?
• source of numerous annotation errors in dbs
• error propagation could lead to an error
  catastrophe
• Further complications arise due to modular nature
  of proteins
• modules are autonomous folding units (protein
  building blocks)
• confer variety of functions on a parent protein,
  by multiple combin-ations of the same module, or
  different modules to form mosaics
• Automatic analysis systems dont distinguish
  orthologues from paralogues dont consider the
  modular nature of proteins

26
(No Transcript)
27

• Monkeys are exploited in different Goldberg
  machines, where they perform different functions
  here, we could not predict a monkey sitting in
  that spot, even with total knowledge of the rest
  of the machine
• Similarity searches are just like this
• identifying the presence of a module tells little
  of the function of the complete system
• knowing most components of a mosaic, we cant
  predict a missing one
• modules (monkeys) in different proteins dont
  always perform exactly the same function

28
The Midnight Zone

• Notwithstanding the lessons of Goldberg machines,
  identifying evolutionary links between sequences
  is useful
• this often implies a shared function
• In the genome era, prediction of function from
  sequence is of more immediate value than is the
  prediction of structure
• However, between distantly-related proteins,
  structure is more conserved than the underlying
  sequences
• thus, some relationships are only apparent at the
  structural level
• Such relationships cant be detected by even the
  most sensitive sequence comparison methods
• the region of identity where sequence comparisons
  fail completely to detect structural similarity
  is the Midnight Zone there is thus a
  theoretical limit to the effectiveness of
  sequence analysis methods

29
Significance

• Appreciating that mathematical biological
  significance are different is crucial it is
  especially important in understanding the
  limitations of
• search alignment algorithms, pattern
  recognition techniques, functional site
  structure prediction tools
• Contrary to popular opinion, there is currently
  still
• no biologically-reliable automatic multiple
  alignment algorithm
• no infallible pattern-recognition technique
• no reliable gene, function or structure
  prediction algorithms

30
(No Transcript)
31
Computers dont do biology!
32
Biological Databases

• Overview
• Sequence repositories
• SWISS-PROT TrEMBL
• Composite sequence databases
• NRDB, SPTrEMBL
• Family (pattern) resources
• PROSITE, PRINTS, profiles, Pfam, Blocks, eMOTIF
• Composite family databases
• InterPro

33
Primary sequence databases

• In the early '80s, when sequence data started to
  accumulate, several labs saw advantages to
  establishing central repositories
• trouble is, many labs. thought this was a good
  idea made their own
• Nucleic Protein
• EMBL PIR
• GenBank SWISS-PROT
• DDBJ MIPS
• JIPID
• TrEMBL
• The proliferation of dbs causes problems
• do they have the same format? Which is the most
  accurate? The most up-to-date? The most
  comprehensive? Which should we use?

34
SWISS-PROT

• Endeavours to provide high-level annotation
• e.g., descriptions of the function of the
  protein, the organisation of its domains, PTMs,
  family disease relationships, variants, etc.
• Contains entries from gt5,000 species
• the bulk of these from just a handful of model
  organisms
• H.sapiens, E.coli, M.musculus, D.melanogaster,
  S.cerevisiae, etc.
• The quality of its annotations sets is apart from
  other dbs
• Consequently, it cannot keep pace with the rate
  of data acquisition from the sequencing centres

35
(No Transcript)
36
(No Transcript)
37
TrEMBL

• A computer-annotated supplement to SP
• has the SP format contains translations of all
  CDSs in EMBL
• It has 2 main sections
• SP-TrEMBL contains all entries that will
  eventually go into SP, but haven't yet been
  manually annotated
• REM-TrEMBL contains sequences not destined to
  be in SP
• Igs, fragments of lt8 residues, synthetic
  sequences, etc.
• Arose from the need for a structured SP-like
  resource, allowing rapid access to genome data,
  without compromising the quality of SP by
  including entries with poor analysis
  insufficient annotation

38
(No Transcript)
39
Composite sequence databases

• A solution to the problem of proliferating dbs is
  to compile a composite
• these render searches very efficient, especially
  if non-redundant
• Trouble is, there are now several composites,
  each with their own format redundancy criteria
  the most commonly used are
• NRDB SPTrEMBL
• PDB SWISS-PROT
• SWISS-PROT TrEMBL
• PIR
• GenPept
• GenPept updates
• NRDB SPTrEMBL are non-identical, not
  non-redundant
• but which is best? Which the most comprehensive?
  The most up-to-date? Which should we use?

40
NRDB

• NRDB is built locally at the NCBI
• it includes weekly updates of SP daily updates
  of GenBank, so is up-to-date comprehensive
• But the simplistic manner of its construction
  causes problems
• multiple copies of the same protein are retained
  as a result of polymorphisms /or sequencing
  errors
• errors corrected in SP are reintroduced when
  retranslated from DNA
• numerous sequences are duplicates of existing
  fragments
• The contents of the db are thus error-prone
  redundant
• NRDB is the default db of the NCBI BLAST service

41
SPTrEMBL

• This resource is intended to be both
  comprehensive minimally redundant
• It contains fewer errors than NRDB, but is not
  truly non-redundant
• 30 of the combined total of SP TrEMBL is
  non-unique
• Further reduction of error rates requires more
  manual intervention better expert db management
  systems

42
Family (pattern) databases

• As well as 1' resources, there are also many
  family or pattern dbs derived from them
• trouble is, they use different 1' sources
  different analysis methods, all have different
  formats!
• But it isn't all bad SWISS-PROT is emerging as
  a standard, most pattern dbs use it as their
  basis
• PROSITE SWISS-PROT Regular expressions
  (patterns)
• PRINTS SWISS-PROT/TrEMBL Aligned motifs
  (fingerprints)
• Pfam SWISS-PROT/TrEMBL Hidden Markov Models
  (HMMs)
• Profiles SWISS-PROT Weight matrices (profiles)
• Blocks InterPro/PRINTS Weighted motifs (blocks)
• eMOTIF Blocks/PRINTS Permissive regular
  expressions

43
Why create pattern databases?

• Pattern dbs arise from the need to make more
  specific functional diagnoses than are possible
  simply by searching the 1's
• They are built on the principle that homologous
  sequences may be gathered together in multiple
  alignments, within which are regions (motifs)
  that show little variation
• these motifs usually reflect some vital
  biological role in terms of either structure or
  function
• Motifs are exploited in different ways to build
  diagnostic patterns for protein families
• new sequences can be searched against dbs of such
  patterns to see if they can be assigned to known
  families
• hence they offer a fast track to the inference of
  function

44
What's in a sequence?
45
Methods for family analysis
Single motif methods
Fuzzy regex (eMOTIF)
Full domain alignment methods
Exact regex (PROSITE)
Profiles (PROFILE LIBRARY)
HMMs (Pfam)
Identity matrices (PRINTS)
Multiple motif methods
Weight matrices (Blocks)
46
The challenge of family analysis

• highly divergent family with single function?
• superfamily with many diverse functional
  families?
• must distinguish if function analysis done in
  silico
• a tough challenge!

47
Know your family
48
The problem with domains
49
PROSITE

• The first pattern db
• based on the idea that a protein family can be
  characterised by a pattern of conserved residues
  within a single motif
• Sequence information in motifs is reduced to
  consensus or regular expressions (regexs) the
  seed regex used to search SP
• results are inspected manually to achieve optimal
  results
• Some families cant be characterised by single
  motifs
• here, additional regexs are created until an
  optimal set is achieved that captures most or all
  of the family
• results are then manually annotated for inclusion
  in the db

50
R-Y-x-DT-W-x-LIVM-ST-T-P-LIVM(3)
51
(No Transcript)
52
(No Transcript)
53
PRINTS

• Most protein families are characterised by gt1
  motif
• it is sensible to use many/all of them to build a
  diagnostic signature
• This is the principle of fingerprints
• these offer improved diagnostic reliability by
  virtue of the biological context provided by
  motif neighbours
• Motifs are excised from alignments by hand
• residue information is augmented via iterative
  searches
• results are manually annotated prior to inclusion
  in the db

54
Motif context
order
1
2
3
4
5
interval
55
(No Transcript)
56
SUMMARY INFORMATION 37 codes involving 8
elements 0 codes involving 7 elements
0 codes involving 6 elements 0 codes
involving 5 elements 0 codes involving 4
elements 1 codes involving 3 elements 0
codes involving 2 elements COMPOSITE
FINGERPRINT INDEX 8 37 37 37 37
37 37 37 37 7 0 0 0 0
0 0 0 0 6 0 0 0 0 0
0 0 0 5 0 0 0 0 0
0 0 0 4 0 0 0 0 0
0 0 0 3 1 0 0 0 1 1
0 0 2 0 0 0 0 0 0 0
0 ----------------------------------------
-- 1 2 3 4 5 6 7 8
True positives.. PRIO_COLGU PRIO_MACFA
PRIO_CEREL PRIO_ODOHE PRIO_GORGO PRIO_PANTR
PRIO_HUMAN O46648 PRIO_SHEEP PRIO_CALJA
PRIO_BOVIN PRP2_BOVIN PRIO_ATEPA PRIO_SAISC
PRIO_PREFR PRIO_PONPY O75942 PRIO_CAPHI
PRIO_CEBAP PRIO_CAMDR PRIO_FELCA PRP1_TRAST
PRIO_RABIT PRP2_TRAST PRIO_PIG
PRIO_CANFA PRIO_CRIGR PRIO_CRIMI Q15216
PRIO_RAT PRIO_CERAE PRIO_MUSPF PRIO_MUSVI
PRIO_MESAU PRIO_MOUSE O46593 PRIO_TRIVU
Subfamily Codes involving 3 elements
Subfamily True positives.. PRIO_CHICK
57
(No Transcript)
58
Profiles Pfam

• An alternative to motif-based methods exploits
  regions between motifs, which also contain
  valuable information
• the full alignment effectively becomes the
  discriminator
• A complex scoring scheme allowing for
  substitutions INDELs is used to create
  family-specific profiles
• These profiles can be used to detect distant
  relation-ships, where only few residues are
  conserved
• this is the basis of the Profile library
• In an extension of this approach, alignments are
  encoded as probabilistic models termed HMMs
• this is the basis of Pfam

59
(No Transcript)
60
(No Transcript)
61
(No Transcript)
62
Blocks eMOTIF

• Various advantages to storing motifs in a raw
  form
• no information is lost, different scoring
  schemes may be used to confer different
  diagnostic potentials on the same data
• Additional dbs have arisen in this way
• Blocks uses families identified in InterPro,
  aligns the sequences detects motifs
  automatically
• BLOCKS-format PRINTS uses motifs in PRINTS with
  the Blocks scoring scheme
• eMOTIF creates permissive regexs from Blocks
  PRINTS
• These dbs are derived fully automatically hence
  offer
• no family annotation (they link back to InterPro
  PRINTS)
• no further family coverage

63
(No Transcript)
64
Composite pattern databases

• To simplify sequence analysis, the family
  databases are being integrated to create a
  unified annotation resource InterPro
• release 4.0 contains 4691 entries
• a central annotation resource, with pointers to
  its satellite dbs
• initial partners were PRINTS, PROSITE, profiles
  Pfam
• new partners include ProDom, TIGRfam, SMART
  hopefully others (e.g., Blocks, MetaFam)
• lags behind its sources
• major role in fly human genome annotation

65
(No Transcript)
66
(No Transcript)
67
Pattern Recognition

• Overview
• Determining significance of db matches
• Pattern recognition methods
• regular expression patterns rules
• fingerprints blocks
• profiles HMMs
• Current status of pattern dbs

68
Pattern recognition methods

• These methods classify proteins into families
• the basis of the methods is multiple sequence
  alignment
• They depend on developing representations of
  conserved elements of alignments that may be
  diagnostic of structure or function, whether from
• homologous sequence families
• sequences that share some structural/functional
  domains

69
Single motif methods
Fuzzy regex (eMOTIF)
Exact regex (PROSITE)
Full domain alignment methods
Profiles (PROFILE LIBRARY)
HMMs (Pfam)
Identity matrices (PRINTS)
Multiple motif methods
Weight matrices (Blocks)
70
Determining significance of database matches

• When searching a db, the challenge for analysis
  methods is to determine if matches are related
  (true-positive) or unrelated (true-negative)
• At a given scoring threshold, it is likely that
  unrelated sequences will be matched erroneously
  (false-positives) some correct matches will be
  missed (false-negative)
• The aim is to improve the resolution between the
  curves - in the overlap, it is difficult or
  impossible to establish if matches are
  significant
• Different methods tackle this problem in
  different ways

71
Resolving true false matches
N
True negative
Score
72
Resolving true false matches
N
True negative
Score
73
Regular expressions (patterns)

• These are derived from single conserved regions
  in alignments
• they are minimal expressions, so sequence
  information is lost
• the more divergent the sequences used, the more
  fuzzy poorly discriminating the regex becomes
• Alignment Regex
• GAVDFIALCDRYF
• GPIDFVCFCERFY G-X-IV-DE-F-IVL-X2-C-DE-R-
  FY2
• GRVEFLNRCDRYY
• Regexs do not tolerate similarity
• sequences either match or not, regardless of how
  similar they are
• matching is a binary on-off event frequently
  misses true matches
• single-motif methods are very hit-or-miss how
  do you know if you've encoded the best region?

74
In the beginning was PROSITE

• G_PROTEIN_RECEPTOR PATTERN
• PS00237
• G-protein coupled receptor signature
• GSTALIVMYWC-GSTANCPDE-EDPKRH-X(2)-LIVMNQGA
  -
• X(2)-LIVMFT-GSTANC-LIVMFYWSTAC-DENH-R
• /TOTAL1121(1121) /POS1057(1057)
  /FALSE_POS64(64)
• /FALSE_NEG112 /PARTIAL48 UNKNOWN0(0)
• This represents an apparent 20 error rate
• the actual rate is probably higher
• Thus, a match to a pattern is not necessarily
  true
• a mis-match is not necessarily false!
• False-negatives are a fundamental limitation to
  this type of pattern matching
• if you don't know what you're looking for, you'll
  never know you missed it!

75
(No Transcript)
76
Regular expressions (rules)

• Regex patterns are most effective when applied to
  highly-conserved, family-specific motifs
• It is often possible to identify, shorter generic
  patterns within sequences, characteristic of
  common functional sites
• Functional site Rule
• N-glycosylation N-P-ST-P
• Protein kinase C phosphorylation ST-X-RK
• Casein kinase II phosphorylation ST-X2-DE
• Such features result from convergence to a common
  property
• glycosylation sites, phosphorylation sites, etc.
• They cannot be used for family diagnosis dont
  discriminate
• they can only be used to suggest whether a
  certain functional site might exist (which must
  then be tested by experiment)
• such patterns are normally termed rules

77
Residue groups for fuzzy regexs

• It is possible to assign residues to groups based
  on various biochemical properties e.g., charge
  size
• using such groups theoretically ensures that
  resulting regexs have sensible biochemical
  interpretations
• small Ala, Gly
• small hydroxyl Ser, Thr
• basic His, Lys, Arg
• aromatic Phe, Tyr, Trp
• aliphatic Val, Leu, Ile, Met
• acidic/amide Asp, Glu, Asn, Gln
• small/polar Ala, Gly, Ser, Thr, Pro
• This is more flexible than exact regex matching

78
Diagnostic limitations

• Consider the sequence motif Asp-Ala-Val-Ile-Asp
  (DAVID)
• results of searching for such a motif will
  differ, depending on the db, the motif length
  whether we use exact or permissive fuzzy regexs
• Pattern Matches
• D-A-V-I-D 71 (99)
• D-A-V-I-DEQN 252
• DEQN-A-V-I-DEQN 925
• DEQN-A-VLI-I-DEQN 2,739
• DEQN-AG-VLI-VLI-DEQN 51,506
• D-A-V-E 1,088 (1,493)
• (number of matches in OWL29.6 ( OWL31.1))
• Use of fuzzy regexs has the potential advantage
  of being able to recognise more distant
  relationships
• the inherent disadvantage that more matches
  will be made by chance, making it difficult to
  separate true matches from noise

79
Fingerprints

• Fingerprints are groups of conserved (ungapped)
  motifs excised from alignments used for
  iterative db searching
• no weighting scheme is used
• searches depend only on residue frequencies
• resulting scoring matrices are thus sparse
• Each motif trawls the db independently
• search results are correlated to determine which
  sequences match all the motifs which match only
  partially
• no information is thrown away
• The iterative process refines the fingerprint
  increases its power
• potency is gained from the mutual context of
  motif neighbours
• results are biologically more meaningful than
  those from single motifs

80
TM domain
TM domain
loop region
81
loop region
TM domain
TM domain
82
A fingerprinting overview
PRINTS
annotation
83
How fingerprints are stored
84

• T C A G N S P F L Y H Q V K D E
  I W R M B X Z
• 0 0 2 0 0 0 0 0 0 7 0 0 1 0 0 0
  0 2 0 0 0 0 0
• 0 0 3 0 0 0 0 0 2 0 0 0 4 0 0 0
  3 0 0 0 0 0 0
• 6 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0
• 1 0 1 1 0 3 0 0 1 0 0 0 3 0 0 0
  0 0 0 2 0 0 0
• 2 0 1 1 0 0 0 0 0 0 0 3 0 4 0 0
  0 0 1 0 0 0 0
• 4 0 1 0 0 1 0 2 0 1 3 0 0 0 0 0
  0 0 0 0 0 0 0
• 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0
  0 0 0 0 0 0 0
• 0 0 0 0 0 6 0 0 0 0 0 0 0 6 0 0
  0 0 0 0 0 0 0
• 0 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0
  0 0 0 0 0 0 0
• 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0
  0 0 10 0 0 0 0
• 9 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0
  0 0 0 0 0 0 0
• 0 0 0 0 0 0 12 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0
• 0 0 4 0 0 2 0 0 6 0 0 0 0 0 0 0
  0 0 0 0 0 0 0
• (b)
• T C A G N S P F L Y H Q V K D E
  I W R M B X Z
• 0 0 4 0 0 0 0 8 4 34 0 0 15 0 0 0
  1 7 0 0 0 0 0
• 0 4 15 0 0 0 0 0 7 0 0 0 37 0 0 0
  10 0 0 0 0 0 0
• 50 0 0 0 0 3 0 18 0 0 0 0 0 0 0 0
  0 0 0 2 0 0 0

• YVTVQHKKLRTPL
• YVTVQHKKLRTPL
• YVTVQHKKLRTPL
• AATMKFKKLRHPL
• AATMKFKKLRHPL
• YIFATTKSLRTPA
• VATLRYKKLRQPL
• YIFGGTKSLRTPA
• WVFSAAKSLRTPS
• WIFSTSKSLRTPS
• YLFSKTKSLQTPA
• YLFTKTKSLQTPA
• (a)
• Key
• (a) motif, with 3 conserved positions
• (b) corresponding frequency matrix
• (c) same matrix, but after 3 iterations
• (d) same matrix, with PAM250 weighting

85
Fingerprint visualisation

• The full potency of fingerprinting is gained from
  the mutual context provided by motif neighbours
• This is important, as the method inherently
  implies a biological context to motifs that are
  matched in the correct order in the query
  sequences, with appropriate distances between
  them
• This allows sequence identification even when
  parts of the fingerprint are absent
• e.g., a sequence that matches only 4 of 7 motifs
  may still be diagnosed as a true match if the
  pattern of motif matching is consistent with that
  expected of true neighbouring motifs
• Such matches are best visualised graphically

86
Visualising fingerprints
ID
PRINTS
N
C
Query sequence
Missing motif?
87
N
C
88
N
C
89
(No Transcript)
90
(No Transcript)
91
Blocks

• Blocks are groups of motifs derived automatically
  from families identified in InterPro
• sequences are aligned automatically motifs are
  automatically identified by searching for spaced
  residue triplets (e.g., AxxxVxxC)
• a block score is calculated using the BLOSUM62
  matrix
• validity of blocks is confirmed with a 2nd
  motif-finding algorithm
• blocks found by both methods are considered
  reliable
• Sequences within motifs are clustered to reduce
  contributions to residue frequencies from sets of
  closely-related sequences
• each cluster is treated as a single sequence
  given a score that gives a measure of its
  relatedness
• the higher the weight, the more dissimilar the
  segment from others in the block, the most
  distant being given a score of 100
• segments lt80 similar are separated by blank lines

92
(No Transcript)
93
CSC triplet
94
Profiles

• Profiles are scoring tables derived from full
  domain alignments
• these define which residues are allowed at given
  positions
• which positions are conserved which degenerate
• which positions, or regions, can tolerate
  insertions
• the scoring system is intricate, may include
  evolutionary weights, results from structural
  studies, data implicit in the alignment
• variable penalties are specified to weight
  against INDELs occurring in core 2' structure
  elements
• Within a profile, the I M fields contain
  position-specific scores for insert match
  positions
• in conserved regions, INDELs aren't totally
  forbidden, but are strongly impeded by large
  penalties defined in the DEFAULT field
• these are superseded by more permissive values in
  gapped regions
• the inherent complexity of profiles renders them
  highly potent discriminators, but they are
  time-consuming to derive

95
(No Transcript)
96
(No Transcript)
97
Hidden Markov Models

• HMMs are similar in concept to profiles by virtue
  of encoding full domain alignments
• they are probabilistic models consisting of a
  number of inter-connecting states
• essentially, linear chains of match, delete or
  insert states
• Match states are assigned to conserved columns in
  an alignment
• insert states allow for insertions relative to
  match states
• delete states allow match positions to be skipped
• thus, building an HMM from an alignment requires
  each position to be assigned either to match,
  delete or insert states
• HMMs usually perform well, but can be
  over-trained
• they may also suffer if they are created from an
  iterative automatic alignment process if this
  once accepts a false match, the HMM will become
  corrupt

98
An HMM
C
L
Y
E
C
L
W
D
99
Which craft is best?

• The wide variety of methods available leads to
  familiar problems
• which should we use?
• which is the most reliable?
• which is the most comprehensive?
• ......etc.
• None of the pattern-recognition techniques is
  infallible, none of the resulting pattern dbs
  is complete
• bearing in mind the diagnostic strengths
  weaknesses of the different approaches, always
  keeping biological significance in mind, the best
  strategy is simply to use them all

100
Overview of resources

• PROSITE (SIB) - 1108 entries
• single motifs (regexs) - best with small highly
  conserved sites
• Profile library (ISREC) - 300 entries
• weight matrices - good with divergent domains
  superfamilies
• PRINTS (Manchester) - 1750 entries
• multiple motifs (fingerprints) - best for
  families and sub-families
• Pfam (Sanger Centre) - 3071 entries
• HMMs - good with divergent domains
  superfamilies
• InterPro (EBI) - 4691 entries
• derived from PRINTS, PROSITE, Profiles, Pfam,
  ProDom, etc.
• Blocks (FHCRC) - 2608 entries
• multiple motifs (derived from InterPro PRINTS)
• eMOTIF (Stanford)
• permissive regexs (derived from PRINTS BLOCKS)

101
Building a Search Protocol

• Overview
• The usual starting point
• searching the primary data sources
• Pattern recognition methods
• searching the secondary sources
• Structural functional interpretation of results
• Estimating significance
• when do we believe a result?

102
A practical approach

• Given a newly-determined sequence, we want to
  know
• what is my protein?
• to what family does it belong?
• what is its function?
• how can we explain its function in structural
  terms?
• To this end, by searching pattern dbs fold
  libraries, we may recognise patterns that allow
  us to infer relationships with previously-characte
  rised families/folds
• Given the variety of dbs to search, how do we use
  them to build a sensible search protocol for
  novel sequences?

103

• Protein sequence
  database identity search
• e.g., for short fragments, pinpoints
  identical matches
• to probe - may identify correct reading
  frame
• Protein sequence database similarity search
• e.g., nrdb, SPSPTrEMBL - identifies potential
• homologues to probe
• Protein pattern database search
• e.g., PROSITE, profiles, PRINTS, Blocks,
• Pfam - identifies
  family relationships or pin-
• points key
  structural or functional sites
• Known structure No known
  structure
• Structure classification database query
  Protein fold pattern library search
• e.g., scop, CATH, FSSP - provides details
  e.g., threading - identifies compatible
• of structural class, secondary structure
  folds for the probe sequence
• information, ligand-binding, etc.

104
Searching the primary databases

• Identity searching
• the fastest test of an unknown fragment is to
  perform an identity search. This will reveal in
  seconds whether an exact match to the unknown
  peptide already exists
• This can be helpful in identifying the correct
  reading frame following a 6-frame translation
• ccgtactacaactacgctggtgcattcaag
• Forward 0
• PYYNYAGAFK TRFE_XENLA 207 AGIKEHKCSRSNNE
  PYYNYAGAFK CLQDDQGDVAFVKQ
• Forward 1 XLTRSFER 207 AGIKEHKCSRSNNE
  PYYNYAGAFK CLQDDQGDVAFVKQ
• RTTTTLVHS
• Forward 2 TRFE_XENLA TRANSFERRIN PRECURSOR
  - XENOPUS LAEVIS
• VLQLRWCIQ XLTRSFER TRANSFERRIN PRECURSOR
  - XENOPUS LAEVIS
• Reverse 0
• LECTSVVVR
• Reverse 1
• LNAPA!L!Y
• Reverse 2
• !MHQRSCST

105
Similarity searching

• Whether or not an identity search finds a match,
  the next step is to look for similar sequences
• e.g., you may wish to know if a wider family
  exists
• The most rapid simple option is to use BLAST,
  flavours of it, or FastA
• Several features are worthy of note in BLAST
  output
• look for high scores with low P-values (unlikely
  to be random)
• look for clusters of high scores at the top of
  the hitlist (a family?)
• look for trends in the type of sequences matched

106
Ideal results show high scores low E-values
107
(No Transcript)
108
Why bother with pattern searches?

• Primary searches won't always allow outright
  diagnosis
• BLAST FASTA are not infallible
• BLAST, in particular, often can't assign
  significant scores
• results may be complicated by the presence of
  modules, or compositionally-biased regions
• annotations of retrieved hits may be incorrect
• Pattern dbs contain potent descriptors
• so, distant relationships missed by BLAST may be
  captured by one or more of the family or
  functional site distillations

109
(No Transcript)
110
Searching the pattern databases

• Searching PROSITE
• when using PROSITE's Web form, it is advisable to
  exclude rules from the search, otherwise output
  is filled with spurious matches
• results are either match, or no match
• the user has to judge whether hits are significant

111
(No Transcript)
112
(No Transcript)
113
Searching the pattern databases

• Searching Profiles
• the SIB Web server offers access both to profiles
  within PROSITE pre-release (undocumented)
  profiles
• results are highly specific generally
  diagnostically reliable
• if no match is returned, its usually because the
  entry isnt in the db
• matches to undocumented profiles are often
  dead-ends

114
(No Transcript)
115
Searching the pattern databases

• Searching Pfam
• results are returned in HTML tables accompanied
  by simple graphics to illustrate matched domains
• results are specific usually diagnostically
  reliable
• E-values provide the measure of confidence

116
(No Transcript)
117
Searching the pattern databases

• Searching PRINTS
• results are returned in HTML tables on different
  levels
• a best "guess
• the top 10 best-scoring matches
• the raw data
• graphical options provide a visual impression of
  the quality of matches
• results are specific usually diagnostically
  reliable
• combined E- p-values provide the measure of
  confidence

118
(No Transcript)
119
(No Transcript)
120
Searching the pattern databases

• Searching Blocks
• if results of searching PROSITE PRINTS are
  positive, we would expect these to be confirmed
  by searches of the Blocks dbs
• key features to note in the output are
• the description line, the accession codes (which
  indicate which is the matched motif), the
  best-scoring or anchor block
• most important is the detection of multiple block
  hits where this happens, an E-value denotes the
  significance of the match
• single block matches are usually spurious

121
(No Transcript)
122
Searching the pattern databases

• Searching eMOTIF
• as with Blocks, if results of searching PROSITE
  PRINTS are positive, this should be confirmed by
  searches of eMOTIF
• output is given at several stringency levels,
  which indicate the number of false matches to
  expect in the reported results

123
(No Transcript)
124
Which approach is best?

• BLAST frequently fails to assign significant
  scores
• The hit-or-miss nature of single-motif regular
  expressions can render them worthless
• In spite of (because of?) their complexity,
  profiles HMMs are often out-performed by
  simpler motif methods
• The non-weighting system of fingerprints means
  that Twilight relationships may be missed
• The scoring system used to create blocks
  generates large amounts of noise that may obscure
  the signal
• Only PROSITE PRINTS are fully manually
  annotated
• No method alone is best

125
Structural functional interpretation

• Db searches often do little more than identify a
  protein family
• this only scratches the surface we still want
  to know what our protein does what it might
  look like
• The first step is to examine the detailed family
  documentations in PROSITE, PRINTS InterPro
• these should help to elucidate the function of
  the protein
• The next step is to examine the fold
  classification structure summary resources
• e.g., scop, CATH PDBsum, assuming that a
  structure is in fact available.

126
(No Transcript)
127
(No Transcript)
128
(No Transcript)
129
(No Transcript)
130
(No Transcript)
131
Estimating significance

• When do we believe a result?
• a real example.....

132
(No Transcript)
133
(No Transcript)
134
(No Transcript)
135
(No Transcript)
136
(No Transcript)
137
(No Transcript)
138
(No Transcript)
139
(No Transcript)
140
(No Transcript)
141
(No Transcript)
142
(No Transcript)
143
(No Transcript)
144
Conclusions

• What are the lessons for sequence analysis?
• when searching for distant homologues, several
  dbs should be searched
• different methods provide different perspectives
• dbs arent complete their contents dont fully
  overlap
• The more dbs searched, the more difficult it can
  be to interpret results
• hence s/w is being designed to provide
  "intelligent" consensus outputs
• The more computers are involved in automating
  genome annotation, the greater the need for
  collaboration
• especially between s/w developers, annotators
  biologists
• The more data we have to handle, the more
  rigorous we must be in our thinking ( writing)
  if we are to make sense of the complexities
• We are a long way from having reliable tools for
  deducing protein structure function from
  sequence
• but with the right approach, there is hope