statistical testing for MS/MS proteomics bulk samples

Question

[QX]

Hi all,

I have proteomics MS/MS samples for 3 conditions, 3 replicates/conditions. After MaxQuant analysis, I have a count table for where rows are proteins and columns are samples.

I have tried t-test with multiple testing correction but there are no significant hits as the abundance for each condition are very similar. Is this an problems as 3 replicates per sample are very small, where n = 20 is preferable?

As the value for protein are 'abundance', the statistical methods from RNA-seq are not applicable, which are original modeled the read count and not following a normal distribution. Can anyone suggest a statistical method/ package for proteomics data analysis to find the different protein expression in this case?.

Answer 1

[Papyrus]

Often, these and other types data are transformed to more normal-like distributions by taking the log, and then standard methods are used for testing, such as limma. I suggest you take a look at this bioconductor package DEP (which internally uses limma for testing).

[ATpoint]

Seconding this. limma is more powerful than individual t-tests, we use it routinely for proteomics. Be sure that you analyze data on log2 scale and make sure that you impute missing values appropriately. Please see the linked packages, it covers this.