Should the number of significantly DE genes rise with the number of SVs estimated (SVA package)?
0
0
Entering edit mode
18 months ago
Sam ▴ 230

Is there a monotonous relationship between the number of SVs one estimates (sva package), and the number of DE genes one should get as significant? In other words, is it true that the more SVs are used, the more DE genes should be significant?

(I understand that a large number of significantly DE genes is not a positive value in itself; and the more SVs one estimates, the more likely the chance to overfit. Just for understanding's sake )

sva • 708 views
ADD COMMENT
0
Entering edit mode

Could you rephrase your question (and maybe give an example), it's not clear. Are you asking if there is a linear relation between SV number and DE genes number ?

ADD REPLY
0
Entering edit mode

Clarified the question.

ADD REPLY
0
Entering edit mode

No, why should there be?

ADD REPLY
0
Entering edit mode

From SVA manual :

The goal of the sva is to remove all unwanted sources of variation while protecting the contrasts due to the primary variables included in mod. This leads to the identification of features that are consistently different between groups, removing all common sources of latent variation.

Imho, singificantly DE genes are exactly genes that are consistently different between the groups. So, it seems that the general definition of the goal of SVA implies the definition of significantly DE genes.

Also, when I adjust for SVs using limma's removeBatchEffect, and plot the PCA/distance heatmap, the more SVs are included the better the separation between the groups. Even though the statistical method to determine DE genes (via DESeq2) is different from PCA separation, often it is the case that when the separation on the PCA is good, there are many significantly DE genes.

ADD REPLY

Login before adding your answer.

Traffic: 1434 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6