CENTDIST: Discovery of Co-associated Factors by Motif Distribution

Transcription factors (TFs) do not function alone but work together with other TFs (called co-TFs) in a combinatorial fashion to precisely control the transcription of target genes. Mining co-TFs is thus important to understand the mechanism of transcriptional regulation. Although existing methods can identify co-TFs, their accuracy depends heavily on the chosen background model and other parameters such as the enrichment window size and the PWM score cut-off. In this study, we have developed a novel web-based co-motif scanning program called CENTDIST (http://compbio.ddns.comp.nus.edu.sg/~chipseq/centdist/). In comparison to current co-motif scanning programs, CENTDIST does not require the input of any user-specific parameters and background information. Instead, CENTDIST automatically determines the best set of parameters and ranks co-TF motifs based on their distribution around ChIP-seq peaks. We tested CENTDIST on 14 ChIP-seq datasets and found CENTDIST is more accurate than existing methods. In particular, we applied CENTDIST on an Androgen Receptor (AR) ChIP-seq dataset from a prostate cancer cell line and correctly predicted all known co-TFs (8 TFs) of AR in the top 20 hits as well as discovering AP4 as a novel co-TF of AR (which was missed by existing methods). Taken together, CENTDIST, which exploits the imbalanced nature of co-TF binding, is a user-friendly, parameter-less, and powerful predictive web-based program for understanding the mechanism of transcriptional co-regulation. Genome-wide binding analyses of AP4 in LNCaP with DHT (5alpha-dihydrotestosterone) stimulation using ChIP-Seq.

转录因子（Transcription factors, TFs）并非独立发挥功能，而是以组合式调控模式与其他转录因子（称为协同转录因子co-TFs）协同作用，精准调控靶基因的转录。因此，挖掘协同转录因子对于理解转录调控机制至关重要。尽管现有方法可识别协同转录因子，但其准确性高度依赖所选择的背景模型，以及富集窗口大小、位置权重矩阵（Position Weight Matrix, PWM）得分截断值等参数。
本研究开发了一款全新的基于网页的协同基序扫描工具，命名为CENTDIST（访问地址：http://compbio.ddns.comp.nus.edu.sg/~chipseq/centdist/）。与当前已有的协同基序扫描工具相比，CENTDIST无需用户输入任何特定参数与背景信息，而是可自动确定最优参数集，并基于协同转录因子基序在染色质免疫沉淀测序（Chromatin Immunoprecipitation sequencing, ChIP-seq）峰周围的分布情况对其进行排序。
我们在14组ChIP-seq数据集上对CENTDIST进行了测试，结果显示其准确性优于现有方法。特别地，我们将CENTDIST应用于前列腺癌细胞系中的雄激素受体（Androgen Receptor, AR）ChIP-seq数据集，成功在排名前20的候选结果中预测出AR已知的全部8种协同转录因子，同时还发现了AP4作为AR的新型协同转录因子（该结果为现有方法所遗漏）。
综上，CENTDIST利用了协同转录因子结合的不平衡特性，是一款操作简便、无需预设参数且功能强大的基于网页的预测工具，可用于解析转录协同调控机制。本研究同时包含对经5α-二氢睾酮（5alpha-dihydrotestosterone, DHT）刺激的LNCaP细胞中AP4的全基因组结合位点ChIP-seq分析。