臺灣大學: 醫學工程學研究所林文澧;歐陽彥正劉諭承Liu, Yu-ChengYu-ChengLiu2013-03-292018-06-292013-03-292018-06-292010http://ntur.lib.ntu.edu.tw//handle/246246/254853本論文完成了新型以蛋白質序列為基礎之雙模型蛋白質B值預測器。蛋白質結構與蛋白質功能息息相關，蛋白質結構的可變性決定蛋白質功能的類別。因此，從未知結構的蛋白質序列中或者改變蛋白質序列排列組合，蛋白質序列的可變性預測在蛋白質工程上是需要被得到的資訊。佐以其他以蛋白質序列為基礎之蛋白質體特性預測器，未知結構與功能之蛋白質將可被進一步了解，節省了解蛋白質結構與功能的煩瑣工程或合成特定功能的蛋白質所需要的實驗時間。 本論文所提出之新型蛋白質序列可變性預測器，以本研究團隊所研發之QuickRBFR迴歸法為核心，蛋白質的可變性以蛋白質B值為定義依據，蛋白質B值由X光晶體繞射資料取得，從原子電子雲的分佈情形，可估計出原子的在固態環境中的熱能變動情形，進而推測原子在液態環境中的可動性。 相較於其他兩種蛋白質B值預測器，本論文所完成之雙模型蛋白質序列B值預測器，不但在B值預測值與觀察值的相關系數上，表現優於其他兩種蛋白質B值預測器，在B值預估值與觀察值的平均絕對差異上，更改進了過去在高觀察B值與低觀察B值預測不準確的問題，使得本預測器在可變性分類的問題上也獲得了較好的敏感度(Sensitivity)與準確度(F-measure)。同時，本論文提出了蛋白質可變性梯度特徵向量，成功提升了雙模型蛋白質序列B值預測器的預測效能。 蛋白質的可變性可以蛋白質變動的振幅與速率分類，是以蛋白質的可變性定義多形，但其代表的蛋白質運動物理意義不同，結合多種蛋白質序列可變性預測器，可獲得更多的蛋白質可變性資訊，進而可預測蛋白質運動傾向。 關鍵字 : 蛋白質可變性、蛋白質B值、雙模型、蛋白質可變性梯度A novel dual model sequence based protein B-factors predictor has been achieved in this thesis. To perform functions proteins would change their conformations or be triggered by binding ligand or molecular. Therefore, some protein regions should be flexible (adaptive) to adapt interaction with other molecular and some residues should be rigid to accept molecular single. Predicting protein functions from protein sequence could help us to understand functions of proteins, which would be synthesized to possess certain functions or would not be solved structures by tedious experiments. Protein B-factors, generated from X-ray crystallography, are used to be the protein flexibility index in this study. Protein B-factors reflect the thermal-dynamics of protein solid state and then used to presume the movability of proteins in solution. In this study, the dual model sequence based protein B-factors predictor perform a superior performance in correlation coefficient 0.5283, generated from independent test, which compare with other two works. The correlation coefficient measures between experimental B-factors and predicted B-factors, comparing with two other works. Moreover, the fitness of prediction is improved in high B-factor regions and low B-factor regions. In this study, the protein flexibility gradient feature is conducted to improve the prediction performance. There are many different protein sequence based flexibility predictors on variant protein flexibility definitions. Combining information of different protein flexibility predictors may help us to presume protein dynamics. Key works : Protein flexibility, B-factors, Dual model, The flexibility gradient feature2158193 bytesapplication/pdfen-US蛋白質可變性蛋白質B值雙模型蛋白質可變性梯度protein flexibilityB-factorsDual modelFlexibility gradient featurethesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/254853/1/ntu-99-F90548051-1.pdf