In this paper, we present a study of hemoglobin‐hemoglobin interaction with model reduction methods. We begin with a simple spring‐mass system with given parameters (mass and stiffness). With this known system, we compare the mode superposition method with Singular Value Decomposition (SVD) based Principal Component Analysis (PCA). Through PCA we are able to recover the principal direction of this system, namely the model direction. This model direction will be matched with the eigenvector derived from mode superposition analysis. The same technique will be implemented in a much more complicated hemoglobin‐hemoglobin molecule interaction model, in which thousands of atoms in hemoglobin molecules are coupled with tens of thousands of T3 water molecule models. In this model, complex inter‐atomic and inter‐molecular potentials are replaced by nonlinear springs. We employ the same method to get the most significant modes and their frequencies of this complex dynamical system. More complex physical phenomena can then be further studied by these coarse grained models.