Dear List,

Should smooth quantile normalization (SMQ) be used with Limma-Voom if the quantro p-value is <0.05, indicating differences in the distributions? On the one hand, the smooth quantile normalization paper seems to imply this. On the other hand in previous posts Gordon Smyth has stated that TMM is usually the best way to proceed.

Thanks and best wishes,

Rich

Richard Friedman

Herbert Irving Comprehensive Cancer Center Biomedical Informatics Shared Resource

Columbia University Irving Medical Center

llimma-voom has no special relationship to any normalization method, indeed the voom function gives different normalization options.

There no assumption in the limma-voom pipeline that the distribution of logCPM is the same between samples so I don't see the relevance of a p-value test between such distributions.