酵素分類的預測

Abstract

酵素，是催化劑的一種，因其中的化學反應及功能不同，而分成六類．與利用實驗得知相比，利用結構來預測蛋白質功能的方法日益重要．這篇論文中，我們描述了一些以序列及結構為基礎的編碼系統．我們使用不同的編碼系統在兩個方法上，一個是兩階式支持向量機器方法，另一個則是這篇文章中所描述的霍夫曼樹模型的方法．這個利用支持向量機器的霍夫曼樹模型被提供對未知功能的酵素，預測其酵素分類．比較了兩個方法，使用霍夫曼樹模型我們可以得到一個沒有偏倚而且最好可達36%的準確率，這也證實霍夫曼樹模型在酵素分類的預測上是有用的．

Enzymes, as a subclass of catalysts, can be separated into six parts since they have different chemical reactions and protein functions. Methods for predicting protein function from structure are becoming more important than experimental knowledge. In this study, we describe some coding schemes which include both sequence-based and structure-based protein information. We predict the enzyme class for different coding schemes with 2 methods; one is the 2-level SVM model method, one is the Huffman tree model method which is described in this study. This Huffman tree model using support vector machine (SVM) is provided to predict the enzyme classification from the unknown- function enzymes. By comparing with these methods, Huffman tree model is demonstrated useful on enzyme class predicting since we can obtain unbiased and the best prediction accuracy of 36% using the Huffman tree model.