Lipid membranes are crucial components of all cells. The role of cellular membranes is hard to underestimate – they serve as a semi-permeable barrier between cytosol and extracellular environment and accommodate various transmembrane and membrane-anchoring proteins. Lipid bilayers are widely used as a model which effectively captures the main properties of a cell membrane. Molecular simulations are a powerful tool that can be used to gain an insight into complex structure and dynamics of lipid membranes of different composition. In this thesis I explore a possible solution for the problem of efficient sampling of a multicomponent bilayer systems with the Martini model with the use of mixed sampling approaches – MDAS (Molecular Dynamics with Alchemical Steps) and MC-MD (Monte Carlo with Molecular Dynamics). The results suggest that with a mixed sampling scheme one could achieve a significant speed-up of the sampling process. An important role of slowly relaxing long-range electrostatic interactions in the efficiency of MC-MD and MDAS sampling schemes is highlighted. Next, I present a new method for scattering data interpretation for lipid bilayers. The method is based on restrained ensemble molecular dynamics and allows direct incorporation of the scattering data into the simulation via a restraining force that ensures the agreement between simulated and experimental form factors. The resulting model that is in agreement with a given experimental data set can be used as a way to build a structural model that corresponds to the scattering signal. Finally, two applications of molecular simulations of lipids and surfactants are presented. The first one is a combined experimental and computational study of two antimicrobial peptides, myxinidin and WMR in the presence of lipid membranes of different compositions. This study suggests that both peptides actively interact with the negatively charged membranes and localize predominantly in the headgroup region of the bilayer. The presence of the peptides is shown to have a noticeable membrane thinning effect. The multi-microsecond simulations of the peptides reveal the tendency of both peptides to self-aggregate, which is influenced by membrane composition. The second study is focused on the C-terminal FACT domain of ataxia telangiectasia mutated and its interactions with lipid membranes and micelles. NMR structural data, 15 N-relaxation data, NMR data using spin-labeled micelles, and molecular dynamics simulation of the FACT domain suggests that while the structure formed by the three helices of the FACT domain is rather dynamic, none of the helices forms stable contacts with each other. When a membrane mimetic is present, the FACT domain is prone to bind to the membrane or micelle, and all three helices reside predominantly in the headgroup region of the membrane.