Modelling and Assessment of spectroscopic data by Bayesian estimation methods
Rolf Diller – Andreas Wirsen – Jan Hauth – Sandro Keller
The understanding of dynamics in biomembrane systems is experimentally based on measurements, that is, on the collection of data. Current technology in time-resolved absorption or fluorescence spectroscopy allows for measurements over both broad temporal and spectral ranges, each with very high resolution. Accordingly, the resulting datasets are huge, and since the measurements most often do not directly reveal the kinetic models of the processes investigated, further data analysis and inference steps are needed. This includes as well data-reduction techniques and noise removal. The basis for this is the identification of an appropriate mathematical model that describes the experimental data in a physically meaningful context.
Fit algorithms should therefore not only yield appropriate estimates of the parameter values; it is equally important that such algorithms provide information about the reliability of the estimations as well as the goodness of fit of the underlying model structure. Bayesian estimation provides this information in the form of posterior distributions and the marginal data likelihood (sometimes called evidence). The corresponding simulation-based algorithms (also called Monte Carlo methods) are very flexible and mathematically well-grounded, and recent advances in computer technology and algorithmic methodology allow the application of these methods also in scenarios where the models are complex and the amount of data is high.
In the proposed project, this methodology will be further developed, primarily on the basis of datasets arising from spectroscopic investigations on
- spectrally broad-band fluorescence decay measurements on membrane-bound fluorophores (in particular, Laurdan and related membrane probes) in the S. Keller group, with the aim to quantify the dynamics in and hydration properties of native and reconstituted lipid-bilayer membranes in the presence of different integral and peripheral membrane proteins, and
- spectrally broad-band transient absorption measurements in the R. Diller group, exploring properties of bilin-chromophore binding to phycobiliproteins as part of the mechanisms that allow efficient light harvesting and down-hill excitation energy transfer from the exterior rods of the phycobilisomes to the chlorophyll-a of the membrane-bound photosynthetic reaction center.
The computational processing of these data sets including model development and fitting of model parameters will be done in the J. Hauth group.
The results will be important not only for the specific applications that have been considered so far, but the methodology will be applicable in all cases where similar data arise.