Hi there,

I'm comparing RNAseq of an input sample condition with a fraction sample condition derived from that input. My coworker is worried that normalized counts and respective log2 fold-changes out of DESeq2 are influenced by a possible extraction bias on some RNAs in the fraction prep. Therefore it was suggested to assess for changes in expression ranking between the conditions as opposed to log2 fold-changes. As in, genes are first ranked from most expressed to least expressed in both conditions (based on avg. normalized expression out of DESeq2), and then tested for significant increases/decreases in ranks between the conditions.

I haven't been able to find any released method to carry out an assessment of genes ranked by (normalized) expression data. Perhaps I could carry it out on the DESeq objects themselves? I haven't discovered any documentation on this so it must not be possible? If anyone is familiar with rank analysis or has any other insights I would appreciate any advice!

Thank you.

My advice here if a concern is library prep effects is to use RUV or SVA (see our workflow for demo code on how to incorporate this into a DESeq2 analysis).