多标签学习框架下的多效抗菌肽抗菌活性预测及样本评估方法研究

项目名称： 多标签学习框架下的多效抗菌肽抗菌活性预测及样本评估方法研究

项目编号： No.61502074

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 曹隽喆

作者单位： 大连理工大学

项目金额： 20万元

中文摘要： 抗菌肽因其具有高效低毒的广谱抗菌活性且几乎无耐药性问题，被看做是抗生素的最佳替代品，对解决抗生素滥用问题具有重要的意义。对抗菌肽的抗菌活性进行预测能有效帮助了解抗菌肽的作用机理，为抗菌肽药物的设计和改造提供理论依据。目前国内外对于常规抗菌肽的预测成果较多，但在多效抗菌肽的预测和样本评估方面的研究还比较薄弱。本项目拟针对多效抗菌肽，基于生物信息学方法和机器学习技术，构造有效的多肽样本数据集，利用序列编码技术提取多效抗菌肽的融合特征，在新的多标签学习框架下，通过对抗菌肽数据的分布、相似性和标签关联性等信息的有效学习，挖掘多效抗菌活性间的内在作用，探索多效抗菌活性随抗菌肽特征变化的潜在规律，建立合理的多效抗菌肽预测模型和多肽样本评估方法。并根据所提出的理论方法搭建抗菌肽在线预测平台，实现对多效抗菌肽抗菌活性的高精度预测，为探究抗菌肽的特征与多效抗菌活性之间的深层次关联提供研究基础。

中文关键词： 生物信息学；多效抗菌肽；机器学习；预测方法；样本评估

英文摘要： Antimicrobial Peptides (APMs) are seemed as the best substitution of antibiotics because they have high-efficiency，low-toxicity，broad-spectrum antimicrobial activity without drug resistance, and they are very important for solving the problem of antibiotic abuse. Prediction of the antimicrobial activity of APMs can assist to understand the acting mechanism of them and provide theoretical basis for designing and improving APM drugs. So far the domestic and overseas achievements of conventional APMs prediction are abundant, but there are few researches about pleiotropic APMs prediction and sample evaluation. Aiming for pleiotropic APMs and based on bioinformatics and machine learning technology, the present project plans to construct effective peptide sample datasets, extract a fusion feature of pleiotropic APMs by encoding APMs sequences, explore the inner effects among pleiotropic antimicrobial activities and the latent rules of pleiotropic antimicrobial activity changing with APMs features through efficiently learning the information of APMs distribution, similarity as well as label correlation, and establish a whole proper pleiotropic APMs prediction model and a peptide sample evaluation approach under a novel multi-label learning framework. And build an online APMs prediction platform according to the proposed theoretical methods for accurately predicting the antimicrobial activities of pleiotropic APMs, and provide research basis for discovering the relationship between APMs feature and pleiotropic antimicrobial activity.

英文关键词： Bioinformatics;Pleiotropic Antimicrobial Peptide;Machine Learning;Prediction Method;Sample Evaluation