Predicting Protein Interactions by Brownian Dynamics Simulations Abstract: We present a newly adapted Brownian Dynamics (BD) based protein docking method for predicting native protein complexes. The approach includes: global BD conformational sampling, compact complex selection and local energy minimization. In order to reduce the computational costs for energy evaluations, a shell-based grid force field was developed to represent the receptor protein and solvation effects. The performance of this BD protein docking approach has been evaluated on a test set of 24 crystal protein complexes. Reproduction of experimental structures in the test set indicates the adequate configuration sampling and accurate scoring of this BD protein docking approach. Furthermore, we have developed an approach to account the flexibility of proteins, which has been successfully applied to reproduce the experimental complex structure from two unbounded proteins. These results indicate that this adapted BD protein docking approach can be useful for the prediction of protein-protein interactions.