Prediction of Dumbbell-Type Soluble Proteins and Their Center Helices

Dumbbell-type proteins have two domains linked by a long helical segment. Typical examples of dumbbell-type proteins are calmodulin and troponin C, which have regulating function of other proteins [1, 2]. The function seems to be closely related to the dumbbell shape of this protein. Upon binding of calcium, calmodulin changes its structure and strongly interacts with other protein, modulating the function of the bound protein. Troponin C also binds with other proteins, tropomyosin, and regulates the muscle contraction. Therefore, if proteins of dumbbell-type proteins can be predicted from amino acid sequences of total proteome, the method will be useful for inferring the protein-protein interaction and the regulating function of the proteins. In this work, we first analyzed the 3D-structures of proteins in protein data bank (PDB), selecting all dumbbell-type proteins in PDB. Then, we tested various physical parameters of amino acid sequences that may stabilize the dumbbell shape of proteins and constructed an algorithm for predicting dumbbell-type proteins from amino acid sequences alone.