I am trying to figure out if there is a way to modify the design formula such that countData are used as the "effect" / explanatory variable, and another outcome is used as the dependent variable, while adjusting for covariates. So, I'm trying to switch:
countData ~ covariate1 + covariate2 + outcome
to
outcome ~ covariate1 + covariate2 + countData
Is this an option? Or should this kind of question represent something you could still ask about using the first formula listed?
Thank you for your time.
Thank you for the response. This is a more general question for several ongoing projects - we have a large sample count (~500) from human tissue. We are trying to evaluate relationships between gene expression signatures and various epidemiology-based outcomes. For instance, does the expression level of a gene(s) associate with an outcome that occurs later-in-life.
I cannot provide further details through public forum - is that enough?
I wouldn’t usually put an observation like the counts from an RNA-seq experiment (normalized or not) as an independent variable in a design. You may want to have a model that incorporates a latent factor driving the counts and another outcome.
Great- this confirms our thoughts. Thank you for the feedback!
Great- this confirms our thoughts. Thank you for the feedback!