1335The number of three-dimensional (3D) structures of proteins solved by experimental methods is rapidly increasing. As of June 2009, more than 58000 structures were available in the Protein Data Bank (PDB).1) However, the number of amino acid sequences whose 3D structures have not been determined remains more than 100 times greater. Therefore, some approaches for accurate protein structure prediction are urgently required. One of the most effective approaches for protein structure prediction is a comparative modeling (CM) method. This technique uses 3D template structures that have high sequence identities with the target protein. In this paper, we describe our comparative modeling platform which consists of the following four steps: (1) generating and filtering sequence alignments between the target and template proteins; (2) constructing 3D structure models based on each alignment; (3) selecting the best structural model among the candidates; and (4) were primarily used and combined. We have successively participated in the Critical Assessment of Techniques for Protein Structure Prediction (CASP) experiments [8][9][10][11][12] to assess our modeling methods. The CASP experiment is held once every two years with the aim of assessing technical progress in the prediction of protein structures. As a result, protein modeling techniques have progressed. In recent CASP experiments, more than one hundred protein sequences of unknown structures were released by the CASP organizers. The goal for each participating team is to correctly predict the 3D structures from the amino acid sequences. After the prediction period expires, the CASP assessors assess the quality of all models predicted by each participating team. From April to August 2008, the 8th CASP (CASP8) experiment was held and 128 target protein sequences were released.13) We participated in CASP8 as an automatic predictor using the server for which our FAMSD method was internally included. In this paper, the basic algorithm of the FAMSD and the results of the protein modeling for CASP8 targets are described. We show that the modeling method using FAMSD generates reliable 3D protein structures. The protein structures predicted with the FAMSD method should be extremely valuable for experimental researchers. -BLAST, SP3 2) and SPARKS23) were executed. SPARKS2 and SP3 programs were shown to be excellent in the CASP6 experiment.19) Various alignments were obtained and were filtered with our original alignment score value.First, the alignment score value was calculated using the following Eq. 1 for six BLAST-related alignment methods. These methods were BLAST, PSI-BLAST, PSF-BLAST, RPS-BLAST, IMPALA and Pfam-BLAST.Here, Len represents the number of residues of the region aligned to the template protein, SEQid represents the sequence identity percent, SS is the degree of match between the predicted secondary structure elements (SSE) from the target sequence and the SSE of the template protein. The predicted SSE from the amino acid sequence was obtained with PSI-PR...