A large number of genes from many genomes have been determined. Translation of the genes to amino acid sequences and the amino acid analysis of proteins have provided a wealth of protein sequence information. In order to elucidate the biological function of proteins, the three-dimensional (3D) structures of the proteins are essential. The determination of the 3D structure of proteins has been performed with the experimental methods such as X-ray diffraction and nuclear magnetic resonance (NMR) spectroscopy. However, the number of protein structures determined using the experimental methods lags significantly behind the number of protein sequences. Therefore, computational approaches such as comparative or homology modeling for accurate protein structure prediction are urgently required.Consensus methods have been used in the 3D modeling of protein structures.1-3) For example, the reliability of the predicted 3D model is determined by rankings based on factors such as the similarity between Ca atoms of two protein backbones in the compared models. This representative consensus method is very useful if there are numerous accessible modeling or alignment servers. Here, an amino acid sequence of a target protein is used in the analysis of other excellent servers or websites and the output gives the 3D model or the sequence alignment between the target protein and a template protein 3D structure that has been determined experimentally. As shown in the Results and Discussion, we can submit an amino acid sequence under investigation to several web servers and obtain the 3D models or the sequence alignments. Thereupon, the consensus method is an available method, in spite of the complicated way in which we must collect 3D models and alignments created by each web server. Consequently, improving the consensus method is very important in creating a more accurate model based on many 3D models and alignments. In this paper, we have improved the consensus method in taking notice of the environment similarity of side-chains. We explain the algorithm of the new consensus method mentioned above and, actually, use our modeling method employing several server models and sequence alignments collected. As an example of a modeling target, we chose the human Cabin1 protein, a molecule involved in p53 function, which is a very important protein target involved in apoptosis and is combating with cancer in tissues.4-6) Our 3D model for the N-terminal region consisting of 450 residues of human Cabin1 may be useful in the pharmaceutical, medicinal and biological fields.In the meanwhile, 3D-Jury 1) method which is the representative consensus method was one of the most powerful methods to obtain a model with accurate Ca backbone atoms. In the 3D-Jury 1) method each amino acid is represented only by the Ca atom. As such, this method selects "good backbone" protein models that closely match with the experimental structure of the target protein. However, the quality of the side-chains of the selected models is not ensured by this method, beca...