||Protein modeling with molecular mechanics force fields plays an important role in computational biology. However, handling real systems with the current popular all-atom force fields is limited in both size and timescale due to the computationally demanding nature of these force fields. One of the solutions is to use coarse-grained (CG) force field. This thesis reports our effort to develop a new CG force field for biomolecular simulations. To retain the high accuracy that all-atom force fields possess, we use united-atom model to represent each amino acid residue. The solvent model is coarse-grained, with four water molecules represented by one van der Waals particle, so that the simulation speed can be greatly enhanced. Chapter 2 reports the parameterization work for protein-protein and protein-water interactions. Densities and self-solvation free energies of eight pure organic liquids were well reproduced. We could also accurately reproduce the hydration free energies of 105 small organic compounds with an average absolute error of 1.4 kJ/mol. Chapter 3 reports the optimization of backbone and side-chain dihedral parameters and parameterization of protein polar-polar interactions. We show in chapter 4 that this force field is able to fold α-helical, β-sheet and mixed helical/coil peptides. Chapter 5 reports the work for the extension of the force field to study transmembrane peptides. We studied the association of two glycophorin A helices. Our force field gave the best simulated dimer structure so far by current CG force fields. This work demonstrates the high accuracy and applicability of the new force field for biomolecular simulations.