Segmentation Filter

The segmentation filter is useful in cases where you have an existing running quantitation of your genome, but you can see spatially defined regions with consistently different quantitated values. This could be something common such as looking for copy number variation in genomic data, or could be for blocks of enrichment in ChIP, or blocks of methylation in BS-Seq.

The aim of the filter is to divide your existing probes into groups which follow a consistent level of quantitation. Furthermore, the blocks of probes can then be clustered to associate blocks which have similar mean levels of quantitation.

The filter itself uses the fastseg BioConductor package to do the segmentation followed by a custom grouping based on mean values.

Options

1. You must select a single data store whose data will be used to direct the segmentation
2. You can chose an "alpha" value - this is a value between 0 and 1 and determines the propensity of the algorithm to start a new segment. Lower values will tend to produce larger segments, higher values will produce smaller segments. No segment will contain fewer than 4 probes, regardless of the alpha value
3. You can choose whether the segments are defined relative to a change to the previous quantitation (those to the left of the current position), or by setting the global option you can add a constraint that they must also be different to the overall distribution of global values
4. After the initial segmentation you will be shown a graph showing the mean quantitations for the segments defined. On this graph you can add, remove and move break points to say how you want the segments split into subgroups. In addition to the overall list of segments (which will always contain all probes, but with an annotated value to say which segment they fall into) you will also get a probe list for each sub-group defined by the split points set in the second window.