I normally use IGV but loaded some tracks in UCSC. At a particular locus, I did not observe the same thing than in IGV, some peaks disappeared. Looking at ucsc track options, the default is set to a "mean window function". Setting this option to "minimum", "maximum" or "mean+wishker", I can see the same thing than in IGV.
I find that surprising to be the default parameter as, if I did not miss anything, this can be misleading.
Perhaps your track file is too large for all details to display - if I understand correctly, the purpose of windowing function is to facilitate browsing through large number of data points. Personally, I have never noticed peak disappearing - either by comparing igv to ucsc, or changing that setting, I would think anything I consider "peak" would stand out over the background and not get "smoothened" away by changing any of these settings.
Here is the answer from UCSC. My understanding is that the "mean" option, if the binning of your chip-seq bigwig is low, can indeed merge peaks (or make them disappear if a lot of bins have a 0 value). The 'mean+whiskers' is what they seems to advice.
Good Afternoon Nicolas:
I'm assuming you are viewing the graphing type data we
refer to as 'wiggle' data. This type of data can be
as detailed as a single data value for each nucleotide
base in the genome. It depends upon how the data is
constructed to specify how many nucleotide bases a single
data value is associated with. For example, the standard
track in the genome browser 'GC Percent' is calculated at
one data value for five bases in the genome.
Given this density of data values, and depending upon how
large is the view of the genome in the browser, there could
be numerous data values in a single pixel width on the genome
browser image. For example, an 800 wide pixel image in your WEB
browser could be viewing 8,000 nucleotide bases of the genome.
If this was one value per nucleotide base graphing data, then
10 data values are combined to plot the graph of one pixel
wide. The options you are choosing, minimum, maximum, mean,
mean+whiskers is deciding how to combine those 10 data values
to make the graph. The usual case to display the data is 'mean+whiskers'
which will indicate the mean in a darker color, and then one standard
deviation around the mean in a lighter color of all the combined
data values that are used to construct the single pixel wide bit
of the graph. See also:
http://genome.ucsc.edu/goldenPath/help/hgWiggleTrackHelp.html
in particular the 'windowing function' discussion.
