I am learning to use DESeq2 for analyzing time course experiment from RNA-seq workflow: gene-level exploratory analysis and differential expression (section 10) and have a question about the "time" in the analysis. Should the time variable in the fission dataset be considered as factors or numeric values? I worry we would lose information if we consider time points as factors.
Many thanks for your help!
How many time points do you have and how many replicates per time point? Actually by having time as a factor, you allow for more flexible modeling over time (any possible profile over time is accommodated) but you need to have replicates at each time point. When replicates are small or missing, then you have to come up with simple ways to describe how the expression changes over time. I don't recommend ~time with time as a numeric, because this is a simple linear relationship on log counts, and typically expression profiles in time course need something more flexible than this.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Hi Michael, I have 3 time points (0, 20 minutes, 1 hour) and 3 replicates each time point. Would you recommend to use factor or numeric values in this case? Thanks so much!
I'd recommend factors. Factors mean each time point gets its own expected expression value. If you use time as a numeric you need to specify f(time). Choosing f to be linear on log expression is not really justified. So then you have to get into modeling curves, which is certainly possible with DESeq2 and R, but it takes some data exploration and being able to work with the coefficients. So factors are the easy and safe way to go.
Thanks so much your help, Michael!