Detecting distant homologies among protein sequences by comparison of profile hidden Markov models - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Detecting distant homologies among protein sequences by comparison of profile hidden Markov models

Description:

Detecting distant homologies among protein sequences by comparison of profile ... deletes. inserts. matches. The traditional paradigm. HMM of a 'known' family F ... – PowerPoint PPT presentation

Number of Views:43
Avg rating:3.0/5.0
Slides: 12
Provided by: martin353
Category:

less

Transcript and Presenter's Notes

Title: Detecting distant homologies among protein sequences by comparison of profile hidden Markov models


1
  • Detecting distant homologies among protein
    sequences by comparison of profile hidden Markov
    models
  • Martin Madera
  • Department of Computer Science
  • University of Bristol

2
What's a protein? (1/2)?
- linear polymer of amino-acids - protein
sequence string over alphabet of 20 amino-acids
stored in the DNA
sequence of its gene
3
What's a protein? (2/2)?
all atoms
just backbone
cartoon
- proteins fold up into complex 3D structures
- structure functionally important, conserved
during evolution sequences diverge rapidly
- but solving structure experimentally hard, DNA
sequencing easy
4
  • The problem
  • Find if proteins are related
    (homologous)?
  • from their sequences alone
  • Why?
  • Homologs have similar structures and
  • perform similar functions

5
A simple approach
?
MLDQQTINIIKATVPVLKEHGVTITTTFYKNLFAKHPEVRPLFDMGRQES
LEQPKALAM TVLAAAQNIENLPAILPAVKKIAVKHCQAGVAAAHYPIVG
QELLGAIKEVLGDAATDDIL DAWGKAYGVIADVLYAQAVE
PIVDSGSVSPLSDAEKNKIRAAWDLVYKDYEKTGVDILVKFFTGTPAAQA
FFPKFKGLT TADDLKQSSDVRWHAERIINAVNDAVKSMDDTEKMSMKLK
ELSIKHAQSFYVDRQYF AGIIA
- take two sequences - calculate least number of
mutations required to turn one into the other -
compare to numbers for random sequences -
estimate level of significance
6
Profiles of protein sequences
... model a family of related proteins
multiple sequence alignment
7
Profile hidden Markov models (HMMs)?
deletes
inserts
matches
- stochastic machines that emit sequences - each
state emission probabilities transition
probabilities among states
8
The traditional paradigm
?
MLDQQTINIIKATVPVLKEHGVTITTTFYKNLFAKHPEVRPLFDMGRQES
LEQPKALAM TVLAAAQNIENLPAILPAVKKIAVKHCQAGVAAAHYPIVG
QELLGAIKEVLGDAATDDIL DAWGKAYGVIADVLYAQAVE
unknown sequence s
HMM of a known family F
- calculate emission probability P(sHMM), -
compare to null probability P(sNULL), -
calibrate, estimate probability that s belongs to
F
9
An obvious generalization?
?
collect similar unknown sequences, build HMM of
an unknown family A
HMM of a known family F
PRofile Comparer (PRC, http//supfam.org/PRC)? cal
culates
10
Fold recognition results
11
Acknowledgements
Julian Gough, University of Bristol Kevin
Karplus, University of California, Santa Cruz
Write a Comment
User Comments (0)
About PowerShow.com