Question: Signaling in PC3 and PC4
gravatar for shangyuan5000
3 months ago by
shangyuan500010 wrote:

Hi, I have for group conditions and I run PCA analysis in DESeq2. After trying different PCs in the PCA plot, I found that PC3 and PC4 plot could well separate my samples: PCA analysis PC3 and PC4

I think that means my signaling probably mainly located in the PC3&PC4 directions, while PC1 and PC2 are just some other un-relevant things. In this case, is there any tricks to let me focus on the signal I wanted? Or is there any papers discussing this issues? I think this phenomenon is quite common in RNA-seq analysis.

Best regards, Raymond

rna-seq • 214 views
ADD COMMENTlink modified 3 months ago by Kevin Blighe21k • written 3 months ago by shangyuan500010
gravatar for Kevin Blighe
3 months ago by
Kevin Blighe21k
University College London Cancer Institute
Kevin Blighe21k wrote:

Hi Raymond / shangyuan5000

That looks quite convincing. Essentially, PC3 separates M from F, whilst PC4 separates 3 from 4. What do PC1 an PC2 separate?

What you need to do next is to derive the rotated component loadings (or 'variable loadings') to each PC, and then order each of these by absolute value.. The component loadings indicate the strength of each gene to each PC, i.e., the amount of variation that the gene contributes to each PC.

Following my example here ( A: PCA plot from read count matrix from RNA-Seq ), the project.pca object will contain the rotated component loadings data structure, amongst many other structures. Look up the prcomp function to see hw you can access these.


ADD COMMENTlink written 3 months ago by Kevin Blighe21k

Thanks, Kevin. My PC1 and PC2 does not separate the samples by groups, and I have no idea what they are separating since all samples were coming from the same batch as I was told. I will follow your example first and see what's coming out. To look into the loadings seems to be a really good idea. Thanks very much!. PC1 and PC2

ADD REPLYlink written 3 months ago by shangyuan500010
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: 1487 users visited in the last hour