DESeq PCA plot and edgeR MDS plot

Question

[tonja.r]

Both of the tools use different kind of data transformation for doing PCA plots.

However, in the comparison, the plots look quite different, the sample get separated in a different manner. And I have now the same problem.

The MDS plot from edgeR shows almost what I expect (batch effect between the samples) but DESeq PCA plot is different, not showing the expected batch effect.

Why is there such a difference and can different transformations of the data effect the PCA that much?

[karl.stamm]

MDS and pca are different things.

Answer 1

[Lemire]

With MDS, you are trying to reorganize the data points in a reduced dimension space such that the distance between the data points in the reduced space is as similar as possible to the actual distances. But this means that you can rotate the axis, change the signs, etc, without changing the distances. The MDS graph from the link you gave can be rotated to give a graph somewhat similar to the PCA one. I doubt that these were the graphs that are giving you trouble. Maybe you should post yours for a better diagnostic.