A new method for the treatment of holonomic constraints in molecular dynamics simulation has been developed. In this method, constraint forces are solved explicitly. The method explores the special mathematical property (sparse and symmetric positive definite) of an matrix inherent to the molecule under study. The method is especially useful when some bond angles are constrained. A lattice model for proteins is proposed. The connection between implicit-solvent model and explicit-solvent model has been established. The structural features of globular proteins are studied by the model. It is found that the proteins are driven to compact conformations by strong water-water interactions. The constrained molecular dynamics simulations are applied to a short chain in solvent to infer the possible move set for Monte Carlo simulations. It is found that the end monomers should be more mobile than the inner ones and that one and two-monomer moves are the integral part of the set of basic moves. The moves involving four or more monomers can be modeled as the result of several basic moves.