[BioC] Time series analysis: how to tell whether one stands out?

[January Weiner]

Hello, I have two questions concerning the same data set.

I have a number of time series data (say, S1 - S10). Each time series
consists of a few time points (T1-T5), without replicates (you play
what you are dealt...).

First question: there is a hypothesis that one of the time series
stands out in respect to a number of genes. The goal would be to find
these genes.

How can I test whether the single (no repeats) time series for one
gene is different for S1 than S2-S9? I was thinking along this lines:
for one time point, assuming normality of S2-S9, I can calculate the
probability that S1 is from the same distribution as S2-S9. I can then
combine the p-values from the different time points to form a joint
probability, and then correct that one for multiple testing. I think
that this would be even too conservative; is there a better way? Or,
maybe, simply defining the rule that at least one time point should
show a significant adjusted p value?

I can clearly see that certain genes behave differently, and I have a
validation (sort of) in that all genes from one operon behave in a
similar way, e.g. over-expressed at certain time points for S1 but not
S2-S9.

Here comes the next question: is there a standard way of testing the
changes in expression of operons rather than genes? Assuming that I
have a mapping gene->operon in form of a data frame, how can I easily
test whether the expression of an operon differs between the ten time
series?

Best regards,

January

--