Small-angle neutron scattering (SANS) is an important method in many areas of soft matter sciences due to the possibility for contrast variation by hydrogen/deuterium substitution. However, compared to small-angle X-ray scattering (SAXS), SANS remains a less explored method in biological sciences, mainly due to higher sample requirements and more restricted access to neutrons. In this thesis, the overall aim was to develop approaches to apply SANS and contrast variation to study the structure of membrane proteins (MPs) in solution. To be kept in solution, MPs must be stabilized by amphipathic carrier systems, e.g. detergent micelles or phospholipid nanodiscs. Detergent micelles have the advantage of versatile and easy sample handling, but they often destabilize MPs. Nanodiscs have the advantage of providing a lipid bilayer in a well-defined particle, but sample preparation is demanding and not well-understood. Within the major aim, four sub-aims were pursued; (i) understanding the self-assembly of nanodiscs, (ii) developing stability-optimized nanodiscs, (iii) testing size-exclusion chromatography (SEC)-SANS on MP samples, and (iv) developing contrast-optimized detergent micelles and nanodiscs. For objective (i), SAXS was utilized to investigate the structures of nanodiscs formed under different conditions. It was found that nanodisc self-assembly is a fast process independent of detergent removal rate, whereas the lipid stoichiometry and choice of reconstitution detergent are determining factors and important to optimize. In objective (ii), a novel type of nanodisc was developed by introduction of extra negative surface charges together with covalent circularization by sortase A. Combined SAXS, light scattering, circular dichroism spectroscopy and SEC analysis showed that the nanodisc had the expected structure but with drastically increased stability. The bacterial magnesium transport protein CorA could readily be incorporated in these nanodiscs. For objective (iii), a new SEC-SANS setup was benchmarked on a number of nanodisc samples. Despite using relatively dilute samples, data of sufficient quality for structural analysis were obtained. Finally, in objective (iv), uniformly contrast-matched micelles and nanodiscs were utilized to obtain an optimal basis for structural analysis of MPs inside. Using SEC-SANS, high quality SANS data were obtained on CorA. Surprisingly, however, the obtained data could not be explained by current structural models for CorA. On one hand this suggests that more advanced modeling efforts are required, but it also highlights the importance of structural validation by solution-methods, including SAXS and SANS. Overall, the studies of this theses have provided useful insights to the application of SANS to complex biological samples.