Example of SV analyses of a system with non-interacting species. (A) Representative subset of the raw data after elimination of systematic noise contributions. (B) Residuals bitmap from a fit with insufficient quality: the data in (A) are modeled with an impostor single-species fit, resulting in clearly systematic deviations that can be discerned from the strong diagonal feature in the bitmap. (C) The quality of the fit with a c(s) model results in a residuals bitmap with very few diagonal features. There are some vertical and horizontal lines indicative of the remaining residuals due to technical imperfections in the data acquisition process, such as higher-order vibrations of optical components. (D) Size and-shape distribution, transformed into coordinates of sedimentation coefficient and molar mass. The color temperature of the contour lines indicates the population of species. Like in one-dimensional c(s), the peak-width in c(s,M) contains contributions both from regularization (reflecting limited resolution given the signal-to-noise ratio of the data) and from true heterogeneity. (E) Reduction of the c(s,M) distribution to a pure sedimentation coefficient distribution, general c(s,*). This is equivalent to a conventional c(s) analysis but without any constraints to a common average frictional ratio of all species. The inset shows a pure molar-mass distribution, c(M,*), also derived by integration of c(s,M) in a direction orthogonal to c(s,*). (F) Size distribution c(s) using a hydrodynamic scaling law (black line with broad peaks). Also shown is the result of a Bayesian analysis using prior knowledge in the analysis of this non-interacting system, here in the form of c^{(Pδ)}(s) (blue line with sharp peaks) using the hypothesis that the sample consists of discrete species. Generally, the peak width in c(s) can result from either a true polydispersity of the protein (e.g., strong heterogeneity in glycosylation, in conformation, primary sequence, etc.), or from the standard regularization favoring broader peaks for data with low signal/noise ratio. (Figure reproduced from (Schuck et al., 2010).)

## PubMed Commons