Macromolecules, such as RNA and protein, play fundamental roles in all living cells. It is of great importance in both biology and medicine to understand the function of these molecules. A full insight in the molecular function often requires knowledge of the molecular structure in atomic detail. However, obtaining the structure of macromolecules using experimental methods is in many cases difficult and laborious. Consequently, there is a great interest in computational methods for macromolecular structure prediction, as these potentially can become an efficient alternative to the experimental methods. In this thesis I present a number of novel probabilistic methods that address dierent challenges in Markov chain Monte Carlo (MCMC) based structure prediction. These methods are described in four articles. The rst article presents a probabilistic model of RNA conformational space. In the second article, an automated method for estimating generalized ensembles weights in MCMC stimulations is described. The third article introduces a probabilistic model of protein side chains. Finally, the fourth article covers the reference ratio method for combining two distributions describing respectively local and non-local features of macromolecular structures.