A selective view of stochastic inference and modeling problems in nanoscale biophysics (Q1042953)

Advances in nanotechnology of the last two decades enable scientists for the first time to study biological processes on a nanoscale molecule-by-molecule basis. Nanoscale specifically designed single-molecule experiments have many advantages over the traditional experiments involving a population of molecules since single-molecule experiments provide data with more accuracy and higher resolution and give detailed dynamics of a biological process, the type of information rarely available from traditional population experiments. Advances in nanoscale single-molecule biophysics also bring opportunities and challenges for statisticians because the data from single-molecule experiments tend to be much noisier than those from the traditional population experiments and inference of the stochastic dynamics is usually complicated by the presence of latent processes, which are unobserved but affect the data collection. The author considers examples of stochastic inference and modelling problems in this field. He considers likelihood and Bayesian analysis of single-molecule experimental data, outlining, in particular, a Bayesian data augmentation method to handle latent processes and semi-parametric and nonparametric inference of nanoscale data. The second part of the article focuses on stochastic modelling problems. Modelling of subdiffusive motion within single proteins, utilization of stochastic integro-differential equations to model single-protein conformational fluctuations and stochastic networks to model single-enzyme reaction dynamics are discussed.