Normalization on top of TMM scaled data

Question

[weixiaokuan]

Hi

I am trying to use TMM methods to scale my NGS data which has a lot of zeros. After using TMM to scale the data to correct the library size effect, I'd like to apply other normalization approaches and then perform linear/logistic modeling.

I am wondering if it's fine to apply another normalization on top of TMM scaled data? In my understanding, TMM is an effective approach to account for library size. So, after adjusting the data by library size, the following other normalization approaches seem fine.

Let me know your thoughts! Thank you!

-Xiaokuang

[ATpoint]

Linear modelling is usually done on log scale so variance is not a linear function of the magnitude of counts. The cpm function in edgeR can calculate logCPM for you.

[weixiaokuan]

Thank you for your prompt response.

I may rephrase my question that if we may apply other normalization methods on top of TMM scaled data. Is any limitation for additional normalizations?

Thank you!