Is Bayesian Networks Apt To Predict Receptor Conformational States?

Part IIb. Bayesian inference applied to GPCRs in past: predicting receptor-G-protein-coupling


GRIFFIN (G-protein and Receptor Interaction Feature Finding INstrument) is an open source server for the purpose described in the title.


The coupling selectivity is predicted by a support vector machine (SVM) and a hidden Markov model (HMM). The sensitivity in selectivity was reported to be over 85%.

GRIFFIN: a system for predicting GPCR-G-protein coupling selectivity using a support vector machine and a hidden Markov model.
Nucleic Acids Research. 33: Web Server issue W148–W153.

Y Yabuki, T Muramatsu, T Hirokawa, H mukai & M Suwa. 2005.


Short note: There is no sign of update occurred since 2005.


An Overall Flow of Procedure

A query sequence input --> HMM -->; Family determination --> SVM --> Gs coupling
(Non Gs-coupling) --> SVM --> Gi/0 coupling
(Non Gi/0 coupling) --> SVM --> Gq/11 (G12/13?).


A Summary of Featured Elements in SVM Training

1) Lengths of the N-terminal region, all the six loops (intracellular and extracellular), and the C-terminal region.

2) Average hydrophobicity of each transmembrane domain.

3) Bit scores for each query against a sequence profile of amine binding GPCRs, or against a profile of peptide binding GPCRs.

4) Molecular weight of its ligand.

5) Similarity to bovine rhodopsin by having corresponding residues of Pro170 and Lys or Arg at 148 of bovine rhodopsin.

6) Number of Lys or Arg corresponding at positions 244, 247, 248 and 251 of the third intracellular loop (ICL3) in bovine rhodopsin.

7) Number of Lys or Arg corresponding at positions 243, 244, 247, 248 and 251 of ICL3 in bovine rhodopsin.

8) Number of Phe, His, Tyr or Trp present in the ICL3 between the 9th residue and the TM6 interface.

9) Number of Asp or Glu in the ICl3.