Linear Fitting Of Different Patient Samples For The Same Disease Condition

Question

[Sathya]

Hi,

I have 20 samples in my series. In this, 10 are control and 10 are disease (same disease condition) and all are taken from different patient source. After normalizing all the samples, I have done a linear fit which resulted in two columns . One is the coefficients of control and another is the coefficients of disease and using this I have calculated the Fold Change. Is this correct?

The basic doubt is , can I consider all the disease sample as one , linear fitted value ? For getting linear fitted value, I have used lmfit of LIMMA.

[Steve Lianoglou]

It would be helpful if you showed the code that you used for the analysis ... the call from topTable should give you the log2fold change you're looking for (check the logFC column). Also, what's up with the exclamation points in your question's title?!!!!!!!!

[Neilfws]

I've removed the multiple exclamation marks on the basis that looking at them annoys people :)

Answer 1

[Sean Davis]

What you describe is most likely not the appropriate analysis. You'll want to use a contrast matrix to formally test for differences between the two groups. Using a ratio of coefficients defeats the purpose since you will have no measure of significance. As Steve points out, without code, it is difficult to be more directive. That said, be sure you have read and followed the excellent limma user guide for two-sample comparisons.