Structural Bioinformatics 2

1
Structural Bioinformatics (2)

• Dr. D.R. Westhead

2
Reminder of the previous lecture

• Protein 3D structure
• primary, secondary, tertiary and quaternary
  structure
• SSEs and surface loops
• rate of evolution of different parts of protein
  structures
• relevance to multiple sequence alignment

3
Reminder of previous lecture

• Evolution of overall structure
• sequence similarity guarantees structural
  similarity
• 25 sequence ID in an 80 or more residue
  alignment means that the two proteins have the
  same basic 3D structure
• BUT some proteins which share the same structure
  have much lower sequence identities

4
Objectives of this lecture (1)

• To understand the importance of evolutionary
  relationships in prediction problems in
  bioinformatics
• To understand what protein structure prediction
  means
• To understand why protein structure prediction is
  important

5
Objectives of this lecture (2)

• To be aware a number of structure prediction
  methods
• secondary structure prediction
• comparative modelling
• To understand the conditions under which they can
  and should be applied
• To understand their expected accuracy and some
  factors which affect it

6
Prediction in bioinformatics

• Important prediction problems
• prediction of protein sequence from genomic DNA
• prediction of protein function from sequence
• prediction of protein 3D structure from sequence
• prediction of protein function from structure

7
Why are these prediction methods important

• Genome sequencing projects
• generate large quantities of genomic sequences
• BUT what does it mean?
• Prediction of protein sequence, structure and
  function can give clue
• Predictions can be verified experimentally
• often slow

8
Prediction based on similarity

• Many successful prediction methods rely on the
  detection of similarity
• similar sequences have similar functions
• similar sequences have similar structures (last
  lecture)
• For instance doing a BLAST/FASTA search with a
  new protein sequence (previous lectures)

9
The importance of evolution

• Divergent evolution gives rise to many similar
  sequences with related structures and often with
  related functions
• an evolutionary family

10
An evolutionary prediction paradigm

• If a sequence of unknown structure/function can
  be shown to be similar to one or more of known
  structure/function then functional/structural
  details can be transferred to the new sequence.

11
Predicting structure is easier

• As sequences evolve 3D structure tends to be
  conserved
• Function is often conserved as well
• BUT function tends to change with evolution
  faster than structure
• WHY? Or HOW?

12
Protein evolutionary families

• Within an evolutionary protein family we expect
• only one basic 3D structure
• perhaps more than one different function
• Functional differences can be minor or major
• changes in enzyme specificity (minor)
• change from enzyme to structural protein (seen in
  GST family) (major).

13
What is protein structure prediction?

• In its most general form
• a prediction of the (relative) spatial position
  of each atom in the tertiary structure generated
  from knowledge only of the primary structure
  (sequence)

14
Why predict protein structure?

• The sequence structure gap
• 1 000 000 known sequences, 20 000 known
  structures
• Structural knowledge brings understanding of
  function and mechanism of action (last lecture
  and practical)
• Can help in prediction of function

15
Why predict protein structure?

• Predicted structures can be used in structure
  based drug design
• It can help us understand the effects of
  mutations on structure or function
• It is a very interesting scientific problem
• still unsolved in its most general form after
  more than 20 years of effort

16
Methods of structure prediction

• Comparative modelling
• Secondary structure prediction
• Fold recognition/threading
• Ab initio protein folding approaches

17
Terminology

• The (prediction) target sequence
• a sequence of unknown structure for which we
  require a structure prediction

18
Comparative modelling

• Makes a prediction of tertiary structure based on
• sequences of known structure which are similar to
  the target sequence (called template structures)
• an alignment between these and the target
  sequence
• Remember 25 seq ID means two proteins have the
  same basic structure

19
Choice of prediction methods

• If you can find similar sequences of known
  structure then comparative modelling is the best
  way to predict structure
• all other methods are less reliable
• Of course, you cant always find similar
  sequences of known structure.

20
When you cant do comparative modelling?

• The next step is secondary structure prediction
• less detailed results
• only predicts the H (helix), E (extended) or C
  (coil/loop) state of each residue, does not
  predict the full atomic structure
• Example http//www.bioinformatics.leeds.ac.uk/grou
  p/undergraduate/biol3000_blgy2212/ex3_1utg.html

21
Beyond secondary structure prediction

• When you cant do comparative modelling there are
  some things you could do beyond a secondary
  structure prediction
• fold recognition or threading
• ab initio protein folding

22
Fold recognition or threading

• Aimed at detecting when the target sequence
  adopts a known fold, even if it has no
  significant similarity to sequences of known fold
• remember the globin example last lecture
• Beyond the scope of this module

23
Ab initio protein folding

• Aims to predict tertiary structure from basic
  physico-chemical principles
• does not rely on any detection of similarity to
  sequences of known structure
• An important scientific question
• As yet very unreliable for practical predictions

24
Accuracy of structure prediction

• Comparative modelling
• when template and target sequences are closely
  related high accuracy is possible
• sometimes lt 1.0 Angstrom RMSD (RMSD is the root
  mean square deviation between atomic positions in
  predicted and actual structures)
• See handout graph

25
Factors affecting accuracy

• The accuracy of comparative modelling is
  controlled by the quality of the alignment
  between target sequence and template structures
• Alignment is easier if the sequences are closely
  related (e.g. sequence identity gt 80).

26
Accuracy of secondary structure prediction

• The best methods have an average accuracy of just
  about 73 (the percentage of residues predicted
  correctly)

27
Secondary structure prediction methods

• PHD - Rost and Sander (artificial neural network)
• DSC - King and Sternberg (linear discriminant
  analysis)
• NNSSP -Salomov and Solevyev (nearest neighbour
  algorithm)
• PREDATOR - Frishman and Argos
• Jnet Barton Group

28
Use of evolutionary information

• All the previous secondary structure prediction
  methods make use of evolutionary information in
  related sequences
• This improves prediction accuracy enormously
  (without this accuracy would be less than 70)

29
Structure prediction resources

• Secondary structure prediction
• Jpred (http//www.compbio.dundee.ac.uk/Software/JP
  red/jpred.html)
• several others also on the WWW
• Comparative modelling
• SWISSMODEL (http//www.espasy.ch/swissmod/SWISS-MO
  DEL.html)
• We will use these in the practical

30
Caveats

• The prediction methods we have covered are mainly
  aimed at water soluble (globular) proteins
• we still do not know very many 3D structures of
  integral membrane proteins

31
Other aspects of structure prediction

• There are methods for predicting which parts of
  sequences are transmembrane segments
• see for example the ExPASy WWW site
• http//www.expasy.ch/
• There are links to lots of protein prediction
  resources available from ExPASy

32
Summary

• The main points of this lecture were
• prediction using methods based on the detection
  of similarity are important in bioinformatics
• the underlying reason for this is divergent
  evolution of sequences/structures
• prediction of structure by comparative modelling
  is an example of such a method

33
Summary (2)

• Continued
• Comparative modelling should be used to predict
  structures whenever possible
• If it is not possible then secondary structure
  prediction methods can be used, followed by more
  sophisticated methods
• These prediction methods are mainly for globular
  proteins