In performing shell-model calculations for large nuclear systems, the central issue is how to truncate the shell-model space efficiently. It corresponds to a proper arrangement of the configuration space to separate the most important part from the rest of the space. There are different schemes for the shell-model truncation. Considering the fact that most nuclei in the nuclear chart are deformed, using a deformed basis supplemented by angular momentum projection is an efficient way. Shell-model Hamiltonian is then diagonalized in the projected basis. The method is in principle independent of how a deformed basis is prepared and how an effective interaction is chosen. This approach may be viewed as to bridge the two traditional nuclear physics methods: the deformed mean-field approximation and the conventional shell-model diagonalization, because it keeps all the advantages that a mean-field model has to incorporate important correlations, and has the properties of the conventional shell-model that configurations are mixed beyond the mean-filed states to include effects of residual interactions. In this talk, we present the above idea by taking the Projected Shell Model and its extensions as examples [1,2,3,4]. Given the strong demand for shell model calculations also from nuclear astrophysics, one needs such an approach that contains sufficient correlations and can generate wave functions in the laboratory frame, thus allowing exact calculations for transition probabilities, spectroscopic factors, and beta-decay and electron-capture rates, in heavy, deformed nuclei. This research is supported by the National Natural Science Foundation of China (No. 11135005) and by the 973 Program of China (No. 2013CB834401).  K. Hara, Y. Sun, Int. J. Mod. Phys. E4 (1995) 637.  Y. Sun and C.-L. Wu, Phys. Rev. C68 (2003) 024315.  Y. Sun, Int. J. Mod. Phys. E15 (2006) 1695.  Y. Sun, Rev. Mex. Fis. S54(3) (2008) 122.