Poor conformational sampling is the most significant problem in determining stable molecular conformations and free energy changes in molecular systems. This problem is particularly difficult for large molecules such as proteins for which the dimensionality of the conformational space can be of thousands of degrees of freedom. In this work we describe a new method, collective coordinate expansive spaces, for efficient conformational sampling of molecules. The new method performs a dimensional reduction of the molecular system using principal components analysis in dihedral space. The reduced dimensionality representation is subsequently used for conformational sampling by means of a randomized exploration technique. The new method was tested for conformational sampling of four different molecular systems and compared to existing sampling methods. The initial conformational dimensionality of the model systems ranged from 8 to 228 degrees of freedom. The new method displayed the best conformational coverage for systems larger than 36 degrees of freedom.