What Is Moving Average(Ma) In Chip-Chip Data?

Question

[Sunflow]

I have read the paper published on the plant journal"A brassinosteroid transcriptional network revealed by genome-wide identification of BESI target genes in Arabidopsis thaliana",there is a tool "CisGenome" used for ChIP-chip data analysis with the option "MA(moving average)" to find the probable transcription factor binding site, but I don't know what does MA represent, it seems that some statistics must be applied to caculate for subgroups of probes.......

Answer 1

[Chris Evelo]

I haven't done this myself, so I don't know how MA is calculated exactly but conceptually what you say definitely makes sense. ChIP arrays contain many probes for the same promotor region that often are very close together or even strongly overlapping (tiling). What you see happening is that the transcription factor for which you do the ChIP is detected as cross-linked at more than one of these sequences. Now keep in mind that cross-linking is not the same as binding. The transcription factor will most likely really bind at a relatively short sequences but still be very close to the DNA at a longer stretch. So you do get cross-linking over a longer stretch. The best approximating you have for the real binding space is indeed the center of the area where you find the cross-linking weighted for the amount of binding you get and the actual location of the (overlapping) sequences. I would bet that that is precisely what the Moving Average calculates.

Answer 2

[brentp]

As @Chris says, the moving average is utilized because often the probes are overlapping by design--in addition, it's easy to calculate.

There is some discussion of the method and the issues associted with it in this paper: CMARRT: A Tool for the Analysis of ChIP-chip Data from Tiling Arrays by Incorporating the Correlation Structure Sometimes, the moving median is also used as it is more robust against outliers.