[BioC] tricky batch problem

[E. Schwarz]

Dear BioC,

I’m running a metabolomics experiment using samples that were 
collected from years 1999 to 2006. The study is unbalanced as the diseased 
samples were taken in all years, the controls only in year 2000 and 2001. 
The sample size per year varies from 8 to 29 per year.

For some metabolites the measured intensities are significantly different 
between samples from different years in the diseased group. This effect 
(for example a time dependent decay) is neither linear, nor constant for 
all metabolites.

I tried to correct the effect using linear regression if the correlation 
with the time was significant. Due to the nonlinearity of the effect, this 
didn’t solve the problem. Furthermore I’m not sure whether it 
can be assumed that the means of the intensities in all years are equal as 
the number of samples per year is limited (~15).

I’d really appreciate if anyone could suggest a method or a 
BioC-package I could use to correct this effect.

Thanks a lot,

Emanuel