對於轉譯後修飾作用來說，磷酸化反應是一種非常重要的機制。磷酸化反應對於許多的細胞生理作用都有非常大的影響，包括代謝作用、訊息傳遞、細胞分化、以及細胞膜上的物質運輸等等。同時磷酸化反應也被證實與某些疾病有相當大的關係，包括癌症以及阿茲海默氏症等。磷酸化是由於激脢的催化所形成。在不同的激脢催化下，激脢所能辨識的受質結合區也有所不同。在本論文目的是建立電腦分析模型，預測出蛋白質序列中會被磷酸化的胺基酸。本研究運用『隱藏馬可夫模型 (Hidden Markov Model)』來建立模型，主要是依據不同屬性的激脢以及序列組成不同的受質所建立形成的，並將模型實作成預測系統，稱為KinasePhos。使用者在輸入未知磷酸化的蛋白質序列之後，系統會將預測出以何種激脢所催化以及所催化的磷酸化位置。經過與在此之前所提出的研究方法比較，我們有較佳的預測效能。 The phosphorylation of proteins, which is an important mechanism in post-translational modification, affects essentially cellular process such as metabolism, cell signaling, differentiation and membrane transportation. Phosphorylation is performed by protein kinases. The aim here is to computationally predict phosphorylation sites within given protein sequences. The known phosphorylation sites are categorized by substrate sequences and their corresponding protein kinase classes. Profile Hidden Markov Model (HMM) is applied for learning to each group of sequences surrounding to the phosphorylation residues. A predictive tool of protein phosphorylation sites, namely KinasePhos, is implemented to allow users submit protein sequences for prediction of phosphorylation sites. By comparing to other approaches previously developed, our method has higher accuracy and provides not only the location of the phosphorylation sites, but also the corresponding catalytic protein kinases.
