To whom it may concern,

I am currently analysing a ChIP experiment on two different cell lines, tre= ated or not with a drug. To visualize the different binding patterns of the= histone mod I ChIPped, I have first used bamCompare to get the log2ratios = of IP/Input for each sample and without specifying any normalization. Then,= I used computeMatrix in scale regions mode, giving all the bamCompare file= s as input together with a bed file of the common peaks. After this, I visu= alized results on a heatmap with plotHeatmap and have clustered with Kmeans= clustering.

I obtain a certain amount of clusters, then I repeated everything, this tim= e normalizing by RPKM at the bamCompare step. However, the resulting heatma= p and clusters are quite different and I now do not understand which of the= two better represents what my data is actually saying.

I wanted to try a third normalization, by RPGC, with bamCompare, but despit= e this being presented as an option in the bamCompare documentation, whenev= er I run the command an error appears saying "RPGC normalization (--normali= zeUsing RPGC) is not supported with bamCompare!".

Can anyone suggest how to proceed? Is it worth trying the RPGC normalizatio= n or should I do something else to establish which normalization is "the ri= ght one" (if such a thing exists) for my data?

Kind regards,

Silvia

Ps using deeptools version 3.3.0