Hi! DiffBind is great stuff!

I am trying to do a DiffBind analysis comparing 6 conditions (CT00, CT04, CT08, CT12, CT16, CT20). I have managed everything until the final Granges output. However, I am trying to understand what the "contrasts" mean. Here is what I got:

12 Samples, 23986 sites in matrix:
     ID Tissue Factor Condition Treatment Replicate Caller Intervals FRiP
1   0-1  Liver      M      CT00    normal         1 counts     23986 0.24
2   0-2  Liver      M      CT00    normal         2 counts     23986 0.24
3   4-1  Liver      M      CT04    normal         1 counts     23986 0.22
4   4-2  Liver      M      CT04    normal         2 counts     23986 0.22
5   8-1  Liver      M      CT08    normal         1 counts     23986 0.22
6   8-2  Liver      M      CT08    normal         2 counts     23986 0.22
7  12-1  Liver      M      CT12    normal         1 counts     23986 0.21
8  12-2  Liver      M      CT12    normal         2 counts     23986 0.21
9  16-1  Liver      M      CT16    normal         1 counts     23986 0.21
10 16-2  Liver      M      CT16    normal         2 counts     23986 0.21
11 20-1  Liver      M      CT20    normal         1 counts     23986 0.17
12 20-2  Liver      M      CT20    normal         2 counts     23986 0.17

21 Contrasts:
   Group1 Members1 Group2 Members2 DB.edgeR
1    CT00        2   CT04        2     6525
2    CT00        2   CT08        2     7440
3    CT00        2   CT12        2     8136
4    CT00        2   CT16        2     7214
5    CT00        2   CT20        2     7918
6    CT00        2  !CT00       10     1246
7    CT04        2   CT08        2     6350
8    CT04        2   CT12        2     7441
9    CT04        2   CT16        2     6923
10   CT04        2   CT20        2     7526
11   CT04        2  !CT04       10      413
12   CT08        2   CT12        2     6630
13   CT08        2   CT16        2     7027
14   CT08        2   CT20        2     7911
15   CT08        2  !CT08       10      744
16   CT12        2   CT16        2     7244
17   CT12        2   CT20        2     8029
18   CT12        2  !CT12       10     2111
19   CT16        2   CT20        2     7432
20   CT16        2  !CT16       10      794
21   CT20        2  !CT20       10     3712

The Granges data output only reported Conc_CT00 and Conc_CT04. So, what do the 21 contrast mean? Should I do one DiffBind analysis for each pairwise comparison (CT00-CT04, CT00-CT08, CT00-CT12...CT16-CT20)?

If you don;t specify contrasts specifically, DiffBind basically attempts every possible contrast, based on the metadata, such that there are at least three samples in each contrast group (you must have specific minMembers=2 to override the requirement for three replicates). As you have six conditions, this includes every condition contrasted with every other condition (15 permutations), plus each condition contrasted against all other conditions (6 more), for 21 possibilities. [Note: in the development version and moving forward, the "not" (!) contrasts are not included by default).

All DiffBind handles are 1:1 contrasts, so what differences do you want to check? you can use dba.contrast() to specify specific contrasts you are interested in by specifying the group1 and group2 parameters. So if you want to see the differences between CT00 and CT08, you could try:

> myDBA <- dba.contrast(myDBA, group1=myDBA$masks$CT00, name1="CT00",
                               group2=myDBA$massk$CT08, name2="CT08")

Hope this helps-

Rory