Hi all,

I have some phosphoproteomics, time-series data that I would like to analyze. I have four conditions (Control, MutA, MutB, MutA+B) and my aim is to i) Identify which phosphosites are behaving differently across time in the double mutatant (MutA+B) when compare to MutA and MutB. Each condition has been measured across 3 time points (1hr, 8hrs, 24hrs), including the control condition. Each timepoint & condition have 4 replicates.

I was reading the limma user's guide but I am unsure about the following:

• is the approach using splines (9.6.2) possible given that I only have 3 time points?
• how to select the correct number of dfs?
• Should I use duplicateCorrelation since the same replicates were measured throught time?
• Is the following design matrix correct in order to answer my question?

Code:

X <- ns(as.numeric(data$time, df =3)
Group <- factor(data$Treatment)
design <- model.matrix(~Group*X)
fit <- lmFit(dat.mat, design)
fit <- eBayes(fit)
Limma.results = topTable(fit,coef = 1,number = Inf)
candidate_genes <- Limma.results %>%
  filter(adj.P.Val < 0.01)

Thank you in advance!

is the approach using splines (9.6.2) possible given that I only have 3 time points?

Not it is not. Your experiment is covered instead by Section 9.6.1 of the User's Guide.

To elaborate a little, with 3 time points you can only estimate 2 independent contrasts between the time points, so trying to estimate a spline curve with 3 independent parameters is futile. In general, you need to have substantially more time points than spline df.

Should I use duplicateCorrelation since the same replicates were measured throught time?

Yes.