A new coarse-grained (CG) model of chitosan has been developed for predicting solution behavior as a function of degree of acetylation (DA). A multiscale modeling approach was used to derive the energetic and geometric parameters of this implicit-solvent, CG model from all-atom simulations of chitosan and chitin molecules in explicit water. The model includes representations of both protonated d-glucosamine (GlcN(+)) and N-acetyl-d-glucosamine (GlcNAc) monomers, where each monomer consists of three CG sites. Chitosan molecules of any molecular weight, DA, and monomer sequence can be built using this new CG model. Discontinuous molecular dynamics simulations of chitosan solutions show increased self-assembly in solution with increasing DA and chitosan concentration. The chitosan solutions form larger percolated networks earlier in time as DA and concentration increase, indicating "gel-like" behavior, which qualitatively matches experimental studies of chitosan gel formation. Increasing DA also results in a greater number of monomer-monomer associations, which has been predicted experimentally based on an increase in the storage modulus of chitosan gels with increasing DA. Our model also gives insight into how the monomer sequence affects self-assembly and the frequency of interaction between different pairs of monomers.