I am Daniel Castillo Secilla, from University of Granada (Spain). I am using cn.mops for detecting CNVs but I have a question about how the paremeters WL (Windows Lentgh) works. When I used samples from the reference human genome hg19 I used a WL very low (5000) because I was trying different sizes, and the resultant plot was right, but then, I tried the same with samples from the current reference genome hg38 and, with this WL the plot has a lot of artifacts as you can see in grch38 with Artifacts image in the link below. So, I setted the WL with the value by default again (25000) and the artifacts of the plot were removed as you can see at figure grch38 without Artifacts image in the link below.
So I don't understand why these artifacts appear when the WL is too small and I hope that you can bring me some light in this situation.
This is an intrinsic problem of the read count data. If you have a small WL, the counts are low and you will easily get large fold-changes: Imagine that most samples of a readcount of 1 and one sample has a read count of 2. Then this would be a strong signal in terms of fold-changes (2-fold change).
Therefore, I suggest both in the paper and the manual to adjust the WL such that on average 100 reads are in a window. This depends of course on your coverage. The function "getReadCountsFromBAM" suggests and appropriate window length if you leave out that parameter.
This information is very useful for me. Thank you very much for your quickly response! I understand your explanation and now I can continue my research without this doubt.
see below