Chi Square Filter

The Chi Square filter is somewhat unusual in that it doesn't use the current quantitations for your data, but instead is an independent test which uses a chi-square test to determine if the ratio of forward to reverse reads underlying each probe in your test list changes significantly between two or more data stores. It was designed originally for use in bisulphite sequencing experiments, but may have applications for other types of experiment.

This test groups together all reads underlying each probe and is therefore prone to bias for probes where read coverage is not even across the probe. For bisulphite analysis its results are therefore complementary to the results you'd get from running the bisulphite methylation over feature pipeline, where each position which is well enough observed is weighed equally in the final value and it's probably a good idea to combine these two types of filtering when assessing these kinds of data.

Options

• You can choose the data stores you want to use for the test. At least 2 stores must be selected.
• You can choose the significance threshold for the analysis
• You can choose whether to apply multiple testing correction to your raw p-values (you really should in nearly all cases)
• If you have quantitated your data as a perecentage you can resample the reads based on the current quantitation. This is ususally used to allow the test to be used after normalising BS-Seq data
• You can choose to test only probes which have a certain level of observation in each data store
• You can choose to filter the results based on the maximum percentile difference between the samples to ensure a certain absolute change, even if a smaller difference would have been significant