Which Methods Do You Prefer To Analysis Microarray Time-Course Data Without Replicates?

Question

[C Shao]

Hi everyone,

I have two microarray data set considering differentiation in two cell lines. 8 time points for each experiment for the control and treatment, and no replicates (sad...).

Now I am trying to find the changed genes with maSigPro, which employes regression model. In one experiment I got 10,000 changed genes, while in the other experiment only less 100 changed genes.

It seems very unreasonable to get such results, I suspect that the lack of replicates may be a problem.

I would like to know your experience with time-course data, and any suggestion for the no-replicates analysis?

Thanks!

Answer 1

[Obi Griffith]

I had a similar situation with expression data for several cell lines, each having 8 time points, and multiple treatments (control, drugA, drugB, drugA+B) but also no replicates. I was looking for treatment effects. I had some luck doing ANOVA for mixed designs (within/between groups). This website recommends the exANOVA stand-alone software and explains the different situations you might want to consider. But, I used the 'ezANOVA' function in the R 'ez' library. When it came down to it, I had a relatively small number of genes of interest so I was able to visualize the data in various different ways to find the patterns of interest.