| Summary: | 碩士 === 國立陽明大學 === 遺傳學研究所 === 91 === In protein classification, the “superfamily” represents the same evolutionary origin. The proteins in the same superfamily may lack significant sequence identities, but they share similar 3D structures, function features and active sites. This thesis focuses on “TIM barrel metallohydrolase (-like) superfamily”. The most common feature in this superfamily is the conserved eight β-strands in the TIM barrel. In addition, a small β domain is observed in some members. The metal-binding site is another conserved signature with four histidines and one aspartate.
In this thesis, an integrated and automatic system for prediction of possible members of this superfamily from the public sequence databases has been developed. This system combines HMM profiles deriving from sequence information of metal binding sites , fold prediction (threading and secondary structure prediction ) and β-domain search. A numeric score is generated for reliability. Some results demonstrate that the accuracy of this prediction system is more than 75%. Some new proteins are also identified and await for proof. In addition, comparisons with other sequence analysis tools revealed the strong prediction power of our system. Further more, it is worth noting that this system has two important features: full automation and prediction of the putative active sites.
Finally, a web server has been established for user uploading sequence for prediction or generating the specific HMM profile. Efficiency of re-modification of our system for prediction of another superfamily is also discussed.
|
|---|
