Yale University
Department of Statistics
Seminar

Monday, October 28th, 2002

Due to the advance of experimental technologies, chemists for the first time are able to follow a (biochemical) reaction on a single molecule basis. At the same time it also raises many statistical problems. In this talk, we will look at the statistical analysis of oneparticular single molecule experiment -- fluorescence lifetime experiment, in which the conformational change of a single DNA hairpin is studied.

The conformational change is modeled as a continuous time Markov chain, which is not directly observable. Instead, the conformational fluctuation details are inferred from changes in photon counts emitted from a dye attached to the DNA molecule, making the Markovian structure hidden. On top of the hidden Markovian structure, the presence of molecule diffusion adds an extra layer of complexity. We illustrate how to write out the likelihood function in a closed-form and how to use a Metropolis-Hastings algorithm and data augmentation methods to compute the posterior distributions of the parameters which describe the detailed conversions of the DNA molecule.

Our method is applied to both simulated data sets and real observations. We will also discuss the problem of model selection in this particular setting.

This work is joint with Haw Yang, Sunney Xie and Jun Liu