DEXseq, several factors influencing splicing

[DerSeb]

Dear all,

I would like to analyze RNAseq data with DEXseq and have questions regarding the model which is most appropriate for my approach.

Basicly, I have a timeline experiment with 5 time points. I expect several isoform transitions in between these timepoints. At the same time, I have three genotypes. Each of them should also have a different splicing pattern at each timepoint. For each sample, there are 3 biological replicates. This means I have 15 samples ( each 3 replicates) in total. I am interested on the influence of the genotype on the splicing transitions over time.


I guess there are several ways to analyse this dataset. When I compare the genotypes at each timepoint individually, I get ~ 50 - 100 differentially spliced genes with some genes occuring at each timepoint.

Now I wonder if I can test all samples at once, obtaining all exons which have a different splicing-timecourse depending on the genotype? How would the formula look like (to estimate dispersion and test for DEU)?

Also, if I consider one genotype to be an intermediate splicing phenotype (full, half, none splicing), how would you incorporate this into the design for the test?

Thanks for any ideas.

All the best,
Sebastian

[Simon Anders]

Basicly, I have a timeline experiment with 5 time points. I expect several isoform transitions in between these timepoints. At the same time, I have three genotypes. Each of them should also have a different splicing pattern at each timepoint. For each sample, there are 3 biological replicates. This means I have 15 samples ( each 3 replicates) in total.

I hope you mean that you have 45 samples, three samples for each genotype--time-point comination. Otherwise, you seem to have a strange concept of "biological replicates".

Now I wonder if I can test all samples at once, obtaining all exons which have a different splicing-timecourse depending on the genotype? How would the formula look like (to estimate dispersion and test for DEU)?

dispersion: count ~ sample + genotype * timepoint * exon

reduced model: count ~ sample + timepoint * exon

full model: count ~ sample + timepoint * exon + timepoint * genotype * I(exon==exonID)

Also, if I consider one genotype to be an intermediate splicing phenotype (full, half, none splicing), how would you incorporate this into the design for the test?

Not even sure what you mean by full and half splicing.

[DerSeb]

Hello Simon, thanks for your fast reply!

Yes, of course I have 45 samples... sorry for the confusion.

About the level of splicing: I am looking at zero, half and full expression of a splicing factor in our system. One could expect a dosage dependent effect on the splicing of regulated exons.