Hi all, I am analyzing 16S sequenced data of human fecal samples. I proccessed my data with qiime and have been using the R package phyloseq for the data analysis. I wish to use the phyloseq to DESeq command of the R package DESeq2 to normalize by data to stabilize the variance, and avoid rarefaction.
My question is that I am not certain of the design I should use. I attach 10 rows of my samples variables info.
#SampleID Treatment Treatment1 Time Sex Age Individual
1.1 P P0 T0 F 33 1
1.2 P P1 T1 F 33 1
2.1 O O0 T0 F 28 2
2.2 O O1 T1 F 28 2
3.1 Control C0 T0 M 24 3
3.2 Control C1 T1 M 24 3
4.1 Control C0 T0 M 28 4
4.2 Control C1 T1 M 28 4
5.1 O+P OP0 T0 M 24 5
5.2 O+P OP1 T1 M 24 5
I had a n=40, which I randomly assigned in 4 groups ( 3 treatments (O, P and the combination of O+P) and a control group). For each group I sequenced a fecal sample prior to the treatment (T0) and after it (T1). So in total I ended up with 80 libraries from 80 samples.
What I want to compare is the difference of composition/abundance 1) between measurements of T0 and T1 within the treatments, and the difference of composition/abundance 2) between the treatments.
At first I used Treatment1 for the design which is a variable that combines the treatment and time. Afterwords I saw in tutorials that people uses those kind of variables separated, and also incorporated patients variable so I used the design ~ Individual + Time + Treatment.
But R throws the error
error in DESeqDataSet(se, design = design, ignoreRank) :
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
this also happens when I put Treatment and Individual in the design, but other combinations like Time and Individual or Treatment and Time, work just fine.
I thought it was good to add the individuals as a variable to the design, considering that the samples that are in the same treatment-time group are the biologica replica, but show great variability in the abundance count (composition of microbiota among individuals are very large in some cases).
I thought that adding the individual variable to the desing would help to account for that variability. However maybe by adding this variable, the degrees of freedom would be bigger and the weight of the Treatment and Time variables explaining the changes in the abundances could become insignificant if the changes are subte?
I have been trying to understand by my self reading the DESeq papers but I lack the statistical knowledge, and would like to ask you for help in understanding the error and what would be the best design for the analysis.
thank you!!! Cheers
My question is that I am not certain of the design I should use. I attach 10 rows of my samples variables info.
#SampleID Treatment Treatment1 Time Sex Age Individual
1.1 P P0 T0 F 33 1
1.2 P P1 T1 F 33 1
2.1 O O0 T0 F 28 2
2.2 O O1 T1 F 28 2
3.1 Control C0 T0 M 24 3
3.2 Control C1 T1 M 24 3
4.1 Control C0 T0 M 28 4
4.2 Control C1 T1 M 28 4
5.1 O+P OP0 T0 M 24 5
5.2 O+P OP1 T1 M 24 5
I had a n=40, which I randomly assigned in 4 groups ( 3 treatments (O, P and the combination of O+P) and a control group). For each group I sequenced a fecal sample prior to the treatment (T0) and after it (T1). So in total I ended up with 80 libraries from 80 samples.
What I want to compare is the difference of composition/abundance 1) between measurements of T0 and T1 within the treatments, and the difference of composition/abundance 2) between the treatments.
At first I used Treatment1 for the design which is a variable that combines the treatment and time. Afterwords I saw in tutorials that people uses those kind of variables separated, and also incorporated patients variable so I used the design ~ Individual + Time + Treatment.
But R throws the error
error in DESeqDataSet(se, design = design, ignoreRank) :
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
this also happens when I put Treatment and Individual in the design, but other combinations like Time and Individual or Treatment and Time, work just fine.
I thought it was good to add the individuals as a variable to the design, considering that the samples that are in the same treatment-time group are the biologica replica, but show great variability in the abundance count (composition of microbiota among individuals are very large in some cases).
I thought that adding the individual variable to the desing would help to account for that variability. However maybe by adding this variable, the degrees of freedom would be bigger and the weight of the Treatment and Time variables explaining the changes in the abundances could become insignificant if the changes are subte?
I have been trying to understand by my self reading the DESeq papers but I lack the statistical knowledge, and would like to ask you for help in understanding the error and what would be the best design for the analysis.
thank you!!! Cheers