Gordon Smyth I am working on differential expression analysis to identify a set of genes that shows significant expression difference between maize lines. Most of the current packages for differential expression analysis are designed to model expression analysis assuming the gene length is the same across samples. But since I am working with genes from two different maize lines, the gene length is not the same across these lines. And therefore, I am looking for robust statistical methods to test for differences in expression that takes into account the bias in gene length.
I learnt that limma-trend could be used but this will only try to partially recover the library sizes. So, what is the best normalization approach I can take for normalizing on both length and library size and then do downstream differential expression analysis? Any help would be appreciated.

Thank you!