Is Empirical Bayes suitable for large samples?

Question

[Kittinun]

The empirical Bayes method is usually used to find differential protein expression in relative quantitative proteomics studies. To my knowledge, this method is suitable for a dataset with small sample size. However, I would like to know if this method is still suitable for a larger dataset (~ 30 samples/group) or if it will lead to a higher false negative rate.

Answer 1

[ATpoint]

You can use that pretty much for any sample size. For small sample size it is pretty much required to gain the necessary power but it is not a malus for large sample sizes. You might stumble over this paper at some point claiming the opposite in a somewhat related context (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02648-4) but it has imo been shown shortly after that their claims are pointless as what they show is simply due to lack of proper QC and presence of unaddressed batch effects

Got an RNA-seq dataset with 50, 100, 200+ samples? Plug it into a differential expression tool and hope for the best? No! You need to consider QC, EDA, and modeling technical variation, or else risk generating spurious results. A thread on papers, methods, and best practices: pic.twitter.com/p7Zn61QjHw

— Michael Love (@mikelove) April 11, 2022

For proteomics you probably want to use limma, and for large sample size in RNA-seq you will probably also use either limma-trend or limma-voom simply because the implemented linear approach scales much better than the approaches implemented in packages such as DESeq2 and edgeR which play out their benefits especially in the presence of limited sample size.