Abstract
20 models were constructed for the comparative modeling section of the Critical Assessment of Fully Automated Structure Prediction-2 (CAFASP2) [1] [2]. Sequence identity between each target and the best possible parent(s) ranged between 6% and 52%. Searching for the reference proteins and sequence alignments between the targets and reference proteins was provided by PSI-BLAST[3]. The modeling protocol was executed by the automated computer software FAMS[4], consisting of a database search and simulated annealing. There was no human intervention in checking the process of sequence alignments and building models. Both our team and another team, which used 3D-JIGSAW[5], succeeded in solving eight target models. The accuracies of the modeled backbone and side chains were estimated by using torsion angles. In particular, our modeled side chains were significantly more accurate than the ones modeled by the JIGSAW team. Moreover, our backbone models were also better than those of the JIGSAW team.
