Tools for Protein Informatics

1
Tools for Protein Informatics

• Proteomics qualitative and quantitative
  comparison of a proteome (complete set of
  proteins of an organism) under different
  conditions to unravel biological processes.
• Sequence and structure comparison
• Multiple alignments / phylogenetic trees
• composition/ pI/ mass analysis
• Motif and pattern identification
• Secondary structure prediction, TM prediction,
  etc.
• Threading / Homology Modelling
• Visualization and interpretation

2
Using Multiple Alignments

• Screening for membership in a family or
  superfamily.
• Identification of conserved elements important to
  or defining function.
• Identification of highly variable regions.
• Distinguishing global vs. local patterns of
  similarity characteristic of the structural
  scaffold.
• Create a consensus sequence or profile.
• Determination of the level and sites of
  variability across the members of subgroups /
  families / superfamilies.
• Can confirm distant relationships. If A is
  homologous to B andB is homologous to C, A is
  homologous to C.

3
Multiple Alignment

• Definition
• All multiple alignments are pairwise alignments.
• More.

4
Progressive Alignment Methods

• Feng and Doolittle (1987)
• Compare all sequences pairwise.
• There are N x (N-1) / 2 pairs for N sequences.
• Perform a cluster analysis on the pairwise
  data to form a
• guide tree.
• Build the multiple alignment by first aligning
  the most
• similar pair of sequences, then so on.
• Compare sequence to sequence, sequence to
  alignment,
• or alignment to alignment.

5
Progressive Alignment Methods II

• 1. Alignment to sequence comparisons are
  accomplished by pairwise comparisons.
• Then adjust the position of indels in ALL
  sequences.
• 2. Sequences are aligned in order of decreasing
  similarity.
• Multiple algorithms for construction of guide
  tree.
• (i.e. neighbor joining clustering method)
• Sequences are weighted according to their
• relatedness from the guide tree.

6
Progressive Alignment Methods III

• The substitution matrix can be chosen on the fly
• according to the similarity of the guide
  tree.
• 6. Once a group has been aligned, one should
  use the position specific information from the
  multiple alignment to align the next sequence.
• 7. The amount of sequence conservation and
  gapping at any position along the alignment is
  used to help determine mismatch penalties and gap
  opening and extension penalties. Varies
  parameters at different positions, different
  cycles.

7
Gap Penalties

• Lower the gap penalties at positions where gaps
  already occur
• Increase gap penalties adjacent to positions
  where gaps already occur.
• Reduce gap penalties where stretches of
  hydrophilic residues occur.
• Increase or decrease gap penalties using tables
  of the observed frequencies of gaps adjacent to
  each of the 20 amino acids.

8
Profiles

• Information in a multiple alignment is
  represented quantitatively as a table of position
  -specific symbol comparison values and gap
  penalties.
• HHM - Hidden Markov Models -
• probability based models for description of
  alignment.

9
Cautions

• Information OUT directly depends on information
  IN.
• Overrepresentation of a subset of sequences to be
  aligned may bias the inference of an ordered
  series of motifs.
• Multiple alignment is a global alignment.
  Extraneous non-homologous sequence will interrupt
  the alignment of homologous regions.

10
Precomputed Multiple Alignments of Protein
Families

• Pfam http//pfam.wustl.edu/
• HMMs
• Prodom http//protein.toulouse.inra.fr/prodom
  .html
• PSI-Blast
• Systers http//www.dkfz-heidelberg.de/tbi/servic
  es/documentation/systershelp.html
• MetaFam http//metafam.ahc.umn.edu/
• Functional assignments

11
Examples

• ProDom
• Family PD000061 
• WD-REPEAT CONTAINING FACTOR COMPLETE PROTEOME
  TRP-ASP SUBUNIT INITIATION
• http//prodes.toulouse.inra.fr/prodom/2002.1/cgi-b
  in/request.pl?questionDBENqueryPD000061
• Pfam
• Family PF00400 
• WD domain, G-beta repeat
• http//www.sanger.ac.uk/cgi-bin/Pfam/getacc?PF0040
  0

12
Secondary Information from Precomputed Alignments

• Related sequences, related structures, related
  articles, summaries, etc.
• InterPro ProSite CATH
• Pfam Prints SCOP
• ProDom COG MSD
• CD Search PDB
• Blocks Systers
• Homstrad Pandit

13
Applications for Motif Analysis

• Identification of very distant homologs.
• May point to important functional units in a
  protein.
• Can be used to anchor or break-up a multiple
  
• alignment.
• Database of motifs can be used to develop
  other
• informatics applications.

14
ProSite Protein Family Signatures

• http//tw.expasy.org/prosite/
• Signatures include Motifs
• Documentation
• Consensus pattern
• LIVMSTAGC-x(2)-DN- x(2)-LIVMWSTAC-x-LIVMFST
  AG-W-DEN-LIVMFSTAGCN
• Sequences known to belong to the class
• References

15
Meme and Mast

• http//meme.sdsc.edu/meme/website
• Meme motif discovery tool, motif search tool.
• Motifs (highly conserved region) represented as a
  position-dependent letter probability matrices
  which describe the probability of each possible
  letter at each position in the pattern.
• PSSM position-specific scoring matrix.

16
(No Transcript)
17
Protein Sequence Predictions

• Secondary structure
• (alpha helix, beta sheet, turns and coils)
• Transmembrane regions
• Coiled-coil
• Solvent accessible, buried
• Antigenicity
• More

18
Chou Fasman

• Biochemistry (1974) 132, 222.
• Predicts helix, beta, reverse turn, or none.
• Accuracy 77 .
• Based on short-range and medium-range residue
  interactions.
• Training set used to determine secondary
  structure potential of the amino acids.

19
Chou Fasman Amino acid Probabilities
20
Chou Fasman

• Method
• Assign each residue secondary structure
  potential
• (helix, beta, etc.).
• Locate clusters of helix formers, helix
  breakers, etc.
• using a sliding window.
• Search for helical regions.
• Search for beta region.
• Search for turns.

21
GOR

• J. Mol. Bio. (1978) 120, 97.
• J. Mol. Bio. (1987) 195, 957.
• Predicts helix, beta, reverse turn, or coil.
• Accuracy 60 to 70.
• Based on single residue determination rather than
  residue interaction.
• Optimized toward expected percentage secondary
  structure.

22
GOR
23
GOR
24
GOR

• METHOD
• Evaluate the information state for each residue,
  for each conformation state
• Locate the conformation with the highest content
• Add information content from homologous sequence,
  then divide by the number of sequences. Weights
  conserved secondary structural elements.

25
(No Transcript)
26
Accuracy of Predictions

• Information OUT directly depends on information
  IN.
• Inherent errors
• different training sets
• homologous proteins in training set
• different definitions of sceondary structure
• measure of accuracy

27
Recommendations

• Use a number of DIFFERENT programs.
• (statistical versus neural network)
• Use homologous sequences.
• (variable hmologs 90, 80, 70, 60, )
• allow for prediction smoothing.
• (beware end effects)
• Use common sense.

28
(No Transcript)
29
(No Transcript)
30
(No Transcript)