Question: time-series, no replicates, no control groupe
0
17 months ago by
bahlv0
bahlv0 wrote:

Hi,

I'm working with time series RNA-seq data, containig 6 time points. There are no replicates and there is no control groupe. I'm interested to find differential expressed genes between the first timepoint (1/reference) compared to all the other timepoints. The time points can be grouped in 3 groups like this: condition A: 1,2  condition B: 3,4,5  condition C: 6 (but they are not replicates).

I first tried to do a multifactor design with the design (in order to do a likelihood ratio test later):

colData=data.frame(row.names=colnames(ei2),
timepoint=c("1", "2", "3", "4", "5", "6"),
condition=c("A", "A", "B", "B","B", "C"))

design(ddsMF)=formula(~condition+timepoint)

ddsMF=DESeq(ddsMF)

but it gives back the error:

Error in checkFullRank(modelMatrix) :
the model matrix is not full rank, so the model cannot be fit as specified.
One or more variables or interaction terms in the design formula are linear
combinations of the others and must be removed.

I don't know how to fix the problem with the model matrix neither if the approach is right. Are there any suggestions on how to solve this problem or antoher way to find differential expressed genes?

modified 17 months ago by Michael Love26k • written 17 months ago by bahlv0
Answer: time-series, no replicates, no control groupe
0
17 months ago by
Michael Love26k
United States
Michael Love26k wrote:

DESeq2 requires replicates and so you cannot compare each time point to the first one, but you can code time as a numeric and use a smooth function to model the counts. Eg a design of f(time) using a smooth function ‘f’ such as splines or polynomials, with a likelihood ratio test comparing to a reduced design of ~1. You should contact a local statistician to assist with the modeling choices if you have not used splines or polynomial functions in linear models in R before.

Since i'm completly new to DGE analysis, could you give me another advise on how to analyse this dataset? Probably another tool which fits better?

I also thought about comparing each timepoint to the refernce with the help of contrasts (which I know is not the best, since the timepoints are treated as replicates for dispersion estimates)

Above is my best advice for how to analyze the dataset and doesn't really have to do with any specific packages.

Content
Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.