Activity Number:
|
153
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 12, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biometrics Section*
|
Abstract - #301615 |
Title:
|
Bayesian Approach to Parameter Estimation in Individual Protein
|
Author(s):
|
Mikhail Malyutov*+ and Rostislav Protassov and David Golan and Rossen Mirchev
|
Affiliation(s):
|
Northeastern University and Harvard University and Harvard University and Harvard University
|
Address:
|
360 Huntington Ave., Boston, Massachusetts, 02115, USA
|
Keywords:
|
Gibbs sampler ; Individual Protein Tracking ; Markov Chain Monte Carlo ; Method of Moments
|
Abstract:
|
Models for the dynamics of transmembrane and other membrane proteins in intact red blood cells are proposed. A nm-scale gold bead attached to a protein molecule is recorded by our coauthors--experimenters from Harvard medical school--as blurred spot via differential interference contrast icroscopy by a high-speed CCD camera. The centroid position of the bead in each image is determined by a tracking algorithm. The motion of non-transmembrane protein CD-58 is modeled as a sum of three mutually independent components: a Wiener process of some intensity, a Gaussian compound Poisson process due to collisions with other proteins, and white Gaussian noise. We model the motion of transmembrane proteins via a Hidden Markov Chain with the free motion state as above, and the binded state. Model parameters are estimated by the Gibbs sampler and the MCMC using suitable diffuse priors. The uncertainty in parameter values is quantitfied via credible intervals and compared with that obtained by the MOM estimates histogram obtained by fitting each of ten subtrajectories of 1000 positions each. Models' verification is discussed.
|