Question: A weird PCA result using a local Galaxy
gravatar for Gary
9 months ago by
Taiwan/Taichung/China Medical University Hospital
Gary470 wrote:


We use a local Galaxy to run PCA (principal component analysis) based on six mouse RNA-Seq data. However, our result is weird: (1) PC1 can explain nearly all variation (94.5%); (2) All six samples on the 0.4 of PC1. Could you help us? Many thanks.



enter image description here

ADD COMMENTlink modified 9 months ago by Devon Ryan93k • written 9 months ago by Gary470

If almost all the variance is explained by the first PC, it means that the variables are collinear, i.e. they can all be expressed as a linear transformation of one of them. If this is not what you expect, check that the data is really what it should be.

ADD REPLYlink written 9 months ago by Jean-Karim Heriche21k

This is from plotPCA in deepTools, which unfortunately defaults to not transposing the matrix before computing the PCA (I assume it was done this way originally since the PCA() function in matplotlib doesn't accept matrices with more columns than rows). So in this case the results just indicate that "genes are quite variable, but similar between samples", which is OK for basic QC but usually not what people actually care to look at in a PCA.

ADD REPLYlink written 9 months ago by Devon Ryan93k

Many thanks to your super professional answer.

ADD REPLYlink written 9 months ago by Gary470
gravatar for Devon Ryan
9 months ago by
Devon Ryan93k
Freiburg, Germany
Devon Ryan93k wrote:

Make sure to set Transpose Matrix to Yes (it's under Show advanced options). The resulting plot will be much more useful (I wish I'd just made that the default).

ADD COMMENTlink written 9 months ago by Devon Ryan93k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1099 users visited in the last hour