Abstract
Carbohydrate binding modules (CBMs) have been recognized as attaching to carbohydrate-related enzymes to upgrade catalytic efficiency of natural biological polymers including starch and cellulose. Because only a relatively small number of CBM structures have been resolved, alternatively, in silico structures can be established by computational modeling approaches and an accurate target-template sequence alignment is the most deterministic factor for homology modeling. However, the major bottleneck for qualified CBM structure simulation is low sequence identities among CBM members. Fortunately, the conserved characteristics of hydrophilic aromatic residues (HARs) and secondary structure elements were observed and taken into account in feature-incorporated alignment (FIA) to match up the core motifs among CBMs. In this thesis, the improved alignment results were integrated to in silico structure building to increase reliability of predicted structures and to detect potential ligand-binding aromatic residues. The identified HARs and predicted structures deriving from FIA were close to in vitro results and possessed high accuracy and reliability under various criteria. Among FIA and six leading alignment programs, FIA achieved the highest sequence identity and sequence similarity on average with statistically significance. In addition, in silico structures deriving from FIA reported lowest average surface-potential z score with statistically significance. Furthermore, HAR identification can be applied to predict ligand-binding aromatic residues with high true positive rates. Finally, this work comprehensively predicted structures for CBM members and deposited them into a repository named Database of Simulated CBM structures (DS-CBMs). Subsequently, online structure modeling system was developed to allow users to predict their own CBM structures.