TF-coEx:一种基于基因共表达网络的转录因子功能预测新方法

TF-coEx: Transcription Factor Function Prediction based on Gene Co-expression Network

转录因子在细胞内的各种生物通路中起着重要的调控作用.在人基因组中有1 000多个注释为DNA结合蛋白的编码基因,其中部分基因已被证明为转录因子,对它们调控的生物通路也相对比较清楚.其余的大多数DNA结合蛋白可能是潜在的转录因子,但它们的功能并不明确.鉴于转录因子与其所调控的靶基因在基因表达水平上密切关联,本文从基因共表达网络出发建立了1个预测转录因子功能的新方法——co-expression-based transcription factor function prediction(TF-coEx).首先,利用大规模高通量表达芯片数据建立了不同条件下人全基因组的基因共表达网络,并通过网络划分获得包含转录因子的一系列基因共表达模块.之后,通过对模块内基因的功能富集分析,并整合不同网络的模块功能富集结果,对所有潜在的转录因子编码基因进行了功能预测.通过与已知功能的对比,我们证明TF-coEx的预测效果显著好于随机.此外,对预测分值最大的50个结果的文献验证显示,54%的预测有实验证据支持.方法的预测结果为进一步设计具体的实验来验证潜在转录因子的功能提供了方向.

Transcription factors are important for regulating various biological processes.In human genome,more than one thousand genes were annotated to encode transcription factors.Few of them were specified with biological mechanism and DNA-binding motif,the majority not.Here,we are proposing a new computational method-co-expression-based transcription factor function prediction（TF-coEx）-to predict function of transcription factors.In order to cover as much function as possible,we recruited a great deal of microarray data from experiments with multiple conditions.By constructing co-expression networks derived from each microarray data and dividing them into functional modules,we inferred the function of transcription factors based on module function enrichment.After integration the results from all modules,TF-coEx predicted function for all candidate transcription factors,and had a much better performance than random.Moreover,paper validation showed that 54% of the 50 top scored predictions were supported by experimental evidences.Hence,TF-coEx provided guidance for further experiments to unravel function of transcription factors.