摘要
支持向量机有许多优点:有效防止过拟和,适合大的特征空间,给定数据集的信息压缩。本文首次利用支持向量机从氨基酸组成来预测蛋白质的稳定性。总预测率可以达到80.64%,对嗜热蛋白质的预测率为82.50%,对嗜温蛋白质的预测率为80.29%从预测率可以验证氨基酸组成与蛋白质热稳定性成正相关的关系,支持向量机可以成为基于氨基酸组成预测蛋白质热稳定性的有效工具。
Support vector machines have many attractive features, such as effective avoidance of overfitting, the ability to handle large feature spaces, information condensing of the given data set. We firstly use ν-support vector machines to predict protein thermostability from amino acid composition. The total prediction accuracy reaches 80. 64% , the prediction accuracy of thermophilic proteins is 82. 50% , and the prediction accuracy of mesophilic proteins is 80. 29% . From the prediction accuracy, we can conclude that amino acid composition is correlative significantly to protein thermostability, and we regard support vector machines would become a powerful tool for predicting protein thermostability from amino acid composition.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2005年第6期459-465,共7页
Computers and Applied Chemistry
关键词
ν-支持向量机
蛋白质热稳定性
嗜热蛋白质
嗜温蛋白质
v-support vector machines, protein thermostability, mesophilic protein, thermophilic protein
