基于多数据域描述的转录因子结合位点识别  

TRANSCRIPTION FACTOR BINDING SITES RECOGNITION BASED ON MULTIPLE DATA DOMAIN DESCRIPTION

作　　者：陈鸣[1] 薛慧君[2] 熊赟[1] 朱扬勇[1] 

机构地区：[1]复旦大学计算机科学技术学院,上海200433 [2]内蒙古电子信息职业技术学院计算机科学系,内蒙古呼和浩特010011

出　　处：《计算机应用与软件》2011年第6期1-4,42,共5页Computer Applications and Software

基　　金：国家自然基金项目(60903075);上海市重点学科项目(B114)

摘　　要：转录因子结合位点的识别对于理解转录调控机制起着重要作用,也是后基因组时代面临的巨大挑战之一。提出一个基于多任务学习的转录因子位点的识别方法。首先建立一个基于多任务学习理论的多数据域描述模型,然后结合核方法设计转录因子结合位点多分类识别算法。最后对取自于TRANSFAC数据库的真实数据进行交叉验证测试。实验结果表明该方法能充分地利用稀缺的训练样本,有效地捕获不同类别间的联系,从而获得了较高的预测准确率。Transcription factor binding site recognition plays an important role in the comprehension of transcription regulation mechanism,and is the great challenge the post-genome era encounters as well.In this paper we present a multi-task learning based approach for the problem of transcription factor binding sites（TFBS） recognition.Firstly,a new multiple data domain description model was established,it was based on the theory of multi-task learning;then in combination with kernel methods,a multiple classification recognition algorithm for transcription factor binding site was designed.Finally,the real data set retrieved from TRANSFAC database was tested with cross-validation.Experimental result indicated that this approach can fully use scarce training samples and effectually capture the inter-class relations,therefore attained quite high accuracy in prediction.

关 键 词：多任务学习 转录因子结合位点 多数据域描述 核方法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]