Information content and retrieval in solution scattering studies. II. Evaluation of accuracy and resolution

The aim of this paper is to assess the potential accuracy of structure analyses based upon solution X-ray scattering experiments, before the structure problem is solved. In other words, an algorithm is sought expressing accuracy in terms of the experimental observations and associated errors, independently of the value of the structure parameters. The mathematical treatment involves five steps: (a) Data reduction, namely the definition in experimental space of the degrees of freedom of the problem and the determination of their probability law. (b) Choice of representation of the structure consistent with the coarse resolution of solution scattering experiments; a set of sharp pseudo-atoms is adopted, so that the degrees of freedom in real space are the position and electron density of each pseudo-atom. (c) Introduction to a `virtual experiment' defined by a noise distorting the structure parameters, and determination of the corresponding probability law in the experimental space. (d) Comparison of the real and the virtual experiments, or more precisely of the probability laws of the degrees of freedom in experimental space, yielding a relation between the variances of the position and electron density uncertainties, the number of the pseudo-atoms and the value of the structure parameters. (e) Attenuation of the structure dependence via the introduction of a hypothesis of quasi-spherical symmetry, whose effect is to reduce the expression of the variances to a function of the experimental observations, independent of the knowledge of the structure of the particle. This analysis is applied to solution X-ray scattering experiments at variable contrast, interpreted within the framework of the invariant volume hypothesis. The final result takes the form of a curve relating the variances of the position and electron density uncertainties, divided by the number of pseudo-atoms. The example of a low-density serum lipoprotein is used to illustrate these results.