Instruction to using Proximal and Distal (PAD) clustering


Proximal and Distal (PAD) clustering is a web resource to identify and characterize co-localization sites of transcription factor (TF) at regions proximal and distal regions to gene promoters. It builds on top of our finding that proximal and distal binding sites of TFs can facilitate drastically different functions in transcriptional and epigenomic regulations (Oldfield and Yang et al. Mol. Cell 2014). PAD extends such multiple facet TF binding analysis to a large set of TFs profiled by ChIP-Seq technique in embryonic stem cells (ESCs) and leads to the discovery of previously unappreciated function of many more TFs in gene regulation.

Step-by-step guide

Follow these steps to perform analysis:

  1. Selecting the peak files from TFs of interests by clicking the white button. The text field “Selected peaks” shows the peak files that have been selected. A default set of TFs peak files are selected for demonstration purpose.
  2. Selecting the gene annotation files to map the selected peaks for clustering and visualization. Current files are mapped based on RefSeq gene annotation of mm9 assembly. More mapping will be made available in future to facilitate the analysis of ChIP-Seq peak calls of TFs from other cell types and/or organism.
  3. (optional) Upload a user specified peak file for clustering comparison against the TFs selected from the database. This step allows the user to supply and compare their own TF of interest with respect to TFs curated in the PAD database.
  4. Selecting a cut-off value for separating proximal and distal sites. This cut-off value will be used to threshold how close a peak to a TSS annotated in refSeq gene database will be called proximal, and thus separating the binding sites called for each TF into proximal and distal sites. The default value is 1000 bp.
  5. Whether to link the order of the TF clustering heatmaps for proximal and distal sites. If “independent” (the default) is selected, the order of TFs heatmaps for proximal and distal sites will not be linked. Otherwise, the heatmaps will be either linked by the order of proximal sites or distal sites.
  6. (optional) Selecting a p-value cut-off to threshold the heatmaps. If specified, the heatmap cells that does not pass the p-value cut-off will be displayed in white color.

Once the above fields are specified, click submit to run PAD for clustering and visualization.

Interpreting the result

The heatmap visualises the jaccard values between selected (and user-uploaded) peak files from TFs of interest. The dendogram on the top and side of the heatmap shows the clustering of the peak from TFs based on the jaccard values.