The knowledge of protein structures is critical for understanding of molecular machinery, discovering metabolic pathways, rational drug design and many other aspects of life at the molecular level. The number of currently know protein sequences however greatly outnumber known protein structures. Moreover this gap is rapidly growing. Therefore in the past few decades a number of theoretical approaches have been proposed for computational prediction of protein structures. During this project a novel protocol for protein structure prediction will be developed. A new software will be written, tested and applied on a genomic scale. The approach will be based on novel methodological advances, developed during the project. Among the most important elements are: (i) Multiscale approach: during a modeling process it will be possible to switch between several levels of coarse-graining. Low resolution protein representation will be used for an extremely efficient conformational search. Promising regions of solutions space will be further explored by a more detailed and more accurate model. (ii) A new family of force fields will be designed for accurate calculation of protein's energy at each of levels of coarse-graining (i.e. a separate force field for each levels of coarse-graining). (iii) Novel Monte Carlo techniques will be applied and tuned for further speed-up of conformational sampling. Monte Carlo methods will be also applied at the stage of force field development. (iv) Experimental information obtained from SAXS, EM, NMR, mutagenesis and other techniques will be used to direct the search of the conformational space and to increase modeling accuracy. The novel protocol developed during this project allow for structure determination of proteins and protein-protein complexes in a high-throuput fashion. The method will be shared as a publicly available Internet service and tested on representative benchmarks, e.g. a small genome.