基于两层分类器的抗微生物肽种类预测

USING TWO-LAYER CLASSIFIER TO PREDICT THE ANTIMICROBIAL PEPTIDES

抗微生物肽是由宿主产生的一类能够抵御外界病原体感染的小分子多肽,由于其特殊的防御机制和不易产生抗药性,已经成为医学与生物学研究的热点。抗微生物肽应用于医学主要依靠其独特的生物学功能,所以从其功能层面对抗微生物肽进行预测十分必要。本文选取氨基酸组分和伪氨基酸组分信息为特征向量,分别利用随机森林(RF)和k-近邻(KNN)算法,采用两层分类器对抗微生物肽种类进行预测,第一层分类器预测是否为抗微生物肽,成功率达到93.14%,第二层分类器针对抗微生物肽的五种生物学功能进行分类预测,成功率达到83.65%。

Antimicrobial peptides（ AMPs） are a kind of micro- molecule polypeptide defences of most living organisms against invading pathogens. Because they have special defence mechanism,and the resistance of microbes can not easily to be formed,AMPs have become the hot topics in the study of medicine and biology. AMPs used in medicine rely on its unique biological features,so from the aspects of its biological activities to predict AMPs is necessary. In this paper,amino acid composition（ ACC） and Pseudo amino acid composition（ PAAC） were chose as features,two excellent algorithms of Random Forest（ RF） and k- Nearest Neighbors（ KNN） were proposed to predict the antimicrobial peptides by using two- layer of classifier,the first layer of classifier is applied to predict whether a protein is AMP or not,the predictive accuracy is 93. 14%,and the second classifier is proposed to divide AMPs into five groups with diverse biological activities,the best accuracy is 83. 65%.