Potential GPCR Identification with a Quasiperiodic Feature Classifier

1
Potential GPCR Identification with a
Quasi-periodic Feature Classifier

• John Tan
• CSE 598K

2
Overview

• QFC Algorithm
• G-protein Coupled Receptors
• Applications in Drosophila and Anopheles
• Plasmodium and application of the QFC algorithm
• QFC vs. HMMs

3
QFC Algorithm

• Quasi-periodic feature classifier (QFC) relies on
  construction of feature space
• Assumption is that this feature space will
  support interpolation
• A discriminant function is used to optimally
  discriminate between the two classes

4
Training the QFC algorithm

• Training set of 750 GPCRs and 1000 non-GPCRs to
  derive feature space and discriminant function
• 5 parameters to define feature space
• Average periodicity of the hydrophobicity
  function
• Average periodicity of a polarity function
• Variance in the periodicity of the polarity
  function
• Variance in the first derivative of the polarity
  function
• Amino acid usage index

5
Testing the QFC algorithm
6
G-protein Coupled Receptors

• GPCRs are a superfamily of the largest group of
  receptors
• They are always involved in signaling from the
  outside to the inside of the cell where their
  response is felt by a G protein
• Structurally, they have seven hydrophobic
  alpha-helical domains that are usually
  interpreted as seven transmembrane regions

7
GPCRs

• Play an indispensable role in signal transduction
  serving as neurotransmitter, hormone, olfactory,
  and taste receptors
• A substantial proportion of worldwide
  prescription drug sales today are attributed to
  drugs that target GPCRs

8
QFC algorithm applied

• Drosophila melanogaster 2 publications where
  potential odorant and taste receptors were
  identified
• RT-PCR to determine tissue-specific expression
  look for expression specific to the corresponding
  sensory organ

9
(No Transcript)
10
QFC algorithm applied

• Anopheles gambiae 1 publication identifying
  chemosensory receptors. RT-PCR to determine
  tissue-specific expression
• It is likely that receptors are uniquely
  important to the species that they operate in
  e.g. receptors for fruit odors in Drosophila and
  human odors for Anopheles

11
Plasmodium falciparum

• Parasite that causes malaria
• Huge burden on affected areas annually causes
  over 300 million cases and over 1 million deaths
• Disease is spread by mosquitoes. They are an
  integral part of the disease cycle
• Drug resistant Plasmodium and mosquitoes are now
  rampant

12
GPCRs in Plasmodium

• 1 putative GPCR has been identified
• Potential GPCRs in Plasmodium are intriguing for
  several reasons
• Usually only found in higher organisms
• Attractive as potential drug targets
• What function do they serve?

13
QFC applied to Plasmodium

• Two versions of QFC available, neither of which
  gave results matching those published by the
  original authors

14
Target Validation?

• RT-PCR approaches are not useful parasites are
  internal to the red blood cells
• GFP-fusion protein to determine localization
• Green fluorescent protein (GFP) fluoresces under
  UV light
• Clone coding region for a target gene into a
  vector containing GFP transfect into Plasmodium
  determine localization by fluorescence

15
Target Validation?

• GFP-fusion would be time-consuming
• Focus on top 25 hits tools to predict
  transmembrane regions (TMHMM2) and BLAST
• Of the top 25 hits 14 are hypothetical, 2 are
  pseudogenes, 8 have putative annotation, 1 has
  been assigned a function

16
Target Validation

• All but two have predicted transmembrane domains.
• 12 have gt 7 predicted transmembrane domains
• 11 have lt 7 predicted transmembrane domains
• Genes could represent fragments of the final
  protein product

17
BLAST

• Many hits were to other hypothetical proteins
• Some interesting hits but matches werent highly
  significant
• Example results from the top ranked target
  PFI0400c hypothetical protein with 6 predicted
  transmembrane regions

18
(No Transcript)
19
(No Transcript)
20
Potential GPCRs in Plasmodium

• The one gene already annotated as a putative GPCR
  wasnt identified by any of the algorithms
• If a GPCR was identified, it would be interesting
  to BLAST against the P. falciparum genome to try
  to find other similar genes in the same family

21
QFC vs. HMMs

• QFC algorithm has been changed and HMMs have been
  changed since original publication
• Pfam 4.0 using 4 HMMs as of Nov. 2004, Pfam 16.0
  with 6 HMMs representing GPCRs

22
Discussion and Conclusion

• The QFC algorithm performs at a high level but is
  suited to different uses from HMMs
• HMMs must be constructed and trained from
  sequences of known protein families they will be
  more specific and less generalized
• HMMs are less useful when underlying structure is
  preserved but primary sequence is not as conserved

23
Discussion and Conclusion

• One common theme was the sequence divergence of
  GPCRs in this respect, the QFC is well-suited to
  recognize GPCRs
• Target validation can be a problem
• RT-PCR approaches require knowledge that
  expression will be temporally/spatially specific
  and can be isolated at these points
• BLAST results can be very hit-or-miss

24
Discussion and Conclusion

• The construction of a feature space supporting
  interpolation could be used to create other
  classifiers
• Some protein families may not be able to be
  parameterized with variables that could be used
  to distinguish it from other families
• The QFC algorithm is capable of providing
  guidance in the right direction