多数据域描述转录因子结合位点识别.docVIP

基于多数据域描述的转录因子结合位点识别 陈鸣１　薛慧君2　熊赟１　朱扬勇1 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　呼和浩特 010011) 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。本文提出了一个基于多任务学习的转录因子位点的识别方法。首先建立了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。 多任务学习　转录因子结合位点　多数据域描述　核方法 Transcription Factor Binding Sites Recognition by Multiple Data Domain Description Chen Ming1 Xue Huijun2 Xiong Yun1 Zhu Yangyong1 1(School of Computer Science, Fudan University, Shanghai 200433，China) 2(Department of Computer Science, Inner Mongolia Electronic Information Vocational Technical College, Hohhot 010011，China) This paper presents a multi-task learning approach to the problem of Transcription Factor Binding Sites (TFBS) recognition. Firstly, a new multiple data domain description model was established; it was theoretically founded on the new kernel-based multi-task learning formulation of learning multiple tasks simultaneously in order to capture shared structures among tasks. Then, the model was naturally cast to the case of TFBS recognition with kernel methods. Finally, real data set was retrieved from TRANSFAC database to validate the effectiveness of the proposed method. The experimental result indicated that our multi-task learning approach can significantly improve the prediction accuracy by virtue of using training examples from multiple classes as a whole and meanwhile capturing their inter-class relatedness. Multi-task Learning Transcription Factor Binding Sites Multiple Data Domain Description Kernel Methods  引　言 在分子生物学领域，理解转录调控机制是后基因组时代重大挑战之一。达到这一目标的重要步骤是转录因子结合位点（Transcription Factor Binding Sites: TFBS）的识别。转录因子结合位点是基因上游启动子区域长度为5~15bp的短序列片段，被转录因子结合以调控下游基因。通过生物学试验检测TFBS的方法由于开销大、耗时长等缺点，不适合处理海量数据。因此，越来越多的计算识别方法被提出用于初选待测位点。通常一个转录因子结合位点可以被一个或多个转录因子结合，而相关研究表明这种结合具有较高的特异性。因此，在计算分子生物学领域，TFBS的识别问题可视为一个多类别模式分类问题，即给定一个未知样本，判定它可能被那一类或哪几类转录因子结合。 目前，基于核方法和正则化理论的机器学习分类算法是最常用的方法之一，也是统计学习理论 (Statistical learning theory: SLT)