Interpretation of diffbind coefficients
0
0
Entering edit mode
Alexandre • 0
@095e334e
Last seen 21 hours ago
Hong Kong

I have the following design matrix:

8 Samples, 475298 sites in matrix:
ID Factor Condition Replicate    Reads FRiP
1    ep_293T_siNeg_R1 eprint     siNeg         1  9951758 0.76
2    ep_293T_siNeg_R2 eprint     siNeg         2 13656477 0.77
3    ep_293T_siFUS_R1 eprint     siFUS         1 14988942 0.73
4    ep_293T_siFUS_R2 eprint     siFUS         2 21025467 0.75
5 input_293T_siNeg_R1  input     siNeg         1  8520645 0.78
6 input_293T_siNeg_R2  input     siNeg         2 13948312 0.78
7 input_293T_siFUS_R1  input     siFUS         1 10009573 0.77
8 input_293T_siFUS_R2  input     siFUS         2 10541404 0.77


Factor has 2 leves (eprint,input), Condition has 2 leves (siNeg,siFUS). I then implement the design:

eprint_contrast_add <- dba.contrast(eprint_norm,design = '~Factor+Condition',
reorderMeta=list(Factor="input",Condition="siNeg"))


So reference levels are input and siNeg. Then I add contrasts:

eprint_contrast_add <- dba.contrast(eprint_contrast_add,contrast = c('Condition','siFUS','siNeg'))


My understanding is that in this additive model I'm looking for the overall Condition effect controlling for differences due to Factor (input and eprint). Is that correct ?

I then proceed by adding and retrieving coefficients:

eprint_contrast_add <- dba.contrast(eprint_contrast_add,contrast = c('Condition','siFUS','siNeg'))
eprint_contrast.model.coeffs

> eprint_contrast.model.coeffs
[1] "Intercept"                "Factor_eprint_vs_input"   "Condition_siFUS_vs_siNeg"


I don't fully understand the meaning of these coefficients here, I believe Intercept is the log2 fold mean of siNeg and input (reference levels) and Factor_eprint_vs_input is the difference in log2 fold mean between eprint and input and Condition_siFUS_vs_siNeg being the log2 fold mean difference between siFUS and siNeg. Is this statement correct ?

DiffBind • 87 views