Hello, I have a dataset of n individuals which have been profiled using RNAseq in two conditions, c1 and c2. For each individual, I want to get the logFC between these conditions, after controlling for several covariates that are condition specific (e.g. library size, cell type composition).

At this point, I've controlled for these covariates using limma+voom, leaving me with residuals r1 and r2. Since r2 and r1 have already been log transformed, I believe I can take the logFC as r2-r1. Does this seem reasonable?

Alternately, are there any other approaches you would suggest that might let me maintain the weights from limma+voom?

No, you cannot use residuals to compute logFCs.

From the information you give, I guess that you need to add individual-specific condition effects into the linear model.