DiffBind - dba.report concentration value question
1
0
Entering edit mode
10 weeks ago
brisbio ▴ 30

When you generate a report dataframe with dba.report I understand what each column represents. However with regards to the concentration values (log2 normalised read count) is this value comparable between different DiffBind analyses? For example I carried out DiffBind analysis on transcription factor A comparing 2 different timepoints and a separate one for transcription factor B - but are the concentrations shown in the reports comparable? The reason I ask is becaused it seems that the concentration values in TF B of known regions are the same or higher than in the report for TF A but the read counts for TF A in those known regions are actually higher than B (when you visualise on IGV and known to have higher enrichment in ChIP-qPCR). Is this just a case of the normalisation and therefore you can only compare concentration values within an analysis (i.e regions with higher/lower TF binding between regions) but not really between analyses (i.e. TF B binding in a known region is higher than TF A)? I know this is not really what the DIffBind analysis is for - I am using it to look at changes in TF binding between different timepoints. But I just wanted to clarify more information about how these concentration values are made and what I can use them for.

DiffBind • 188 views
0
Entering edit mode
9 weeks ago
Rory Stark ★ 1.1k

It could very well be down to how the data are normalized in each of the analyses. You can get "raw" (non-normalized) read counts in the reports by setting bNormalized=FALSE when calling dba.report() to compare.

Note that there are a number of subtleties around this such as whether Control reads have been subtracted first (default in previous versions of DiffBind or if there are no greylists), and whether a single model has been used (current version with an active design) or not.