Multi-factor blocking in DiffBind
3
1
Entering edit mode
@jaredandrews07-13809
Last seen 16 months ago

I was wondering if it's possible to account for multiple factors when analyzing samples in DiffBind. I want to block for potential batch effects as well as treatment conditions, but I'm unsure how to do so (or if it's even possible). The docs make it clear how to block for one factor, but don't make any mention of multiple. Hoping Rory will pop in here and clarify.

The relevant portions of my samplesheet (I'd like to block for Treatment and Tissue while using Condition for the actual contrast):

                                     
                                     
                                     
                                     
                                     
                                     

 

diffbind differential binding analysis multiple factor design • 1.1k views
ADD COMMENT
2
Entering edit mode
Rory Stark ★ 4.1k
@rory-stark-5741
Last seen 14 days ago
CRUK, Cambridge, UK

You can specify multiple factors for blocking:

> h3k27ac <- dba.contrast(h3k27ac, categories=DBA_CONDITION, 
                          block=c(DBA_TISSUE,DBA_TREATMENT))

While this may be useful to get a quick look, it probably doesn't do what you really want however. It will create a factor for each unique TISSUE-TREATMENT combination and fit a model for each. It would be more correct to fit a TISSUE model and a TREATMENT model. I'm looking at updating DiffBind to handle factor designs more appropriately, but in the meantime you may want to model this directly in DESeq2 or edgeR.

-Rory

ADD COMMENT
1
Entering edit mode

any update on multi-factor blocking in DiffBind? Ideally I'd like to do the analysis without explicitly resorting to DESEQ2.

ADD REPLY
0
Entering edit mode

Same here ! I would really like it if Diffbind gets updated soon regarding this!

ADD REPLY
0
Entering edit mode
Rory Stark ★ 4.1k
@rory-stark-5741
Last seen 14 days ago
CRUK, Cambridge, UK

The latest release of DiffBind now support unrestricted model designs and contrasts for both edgeR and DESeq2, as well as extensive options for controlling normalization. The updated vignette goes through these changes in detail.

ADD COMMENT

Login before adding your answer.

Traffic: 367 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6