Can I use ballgown count files to feed into limma?

Question

[khf0008]

Hello all, I am doing a project examining differential expression between 4 groups of fruit flies: mated flies on a high calorie diet, virgin flies on a high calorie diet, low calorie mated, and low calorie virgin. Sequencing was done on Illumina single-end 73bp reads. My goal is to be able to compare expression across all groups, and I've been told by a committee member that limma is good for this.

I already have count files from this script. It uses ballgown to create 5 files (e_data.ctab, e2t.ctab, i_data.ctab, i2t.ctab, t_data.ctab) which are described here. To my knowledge, the only type of count data these files contain is FPKM. All the examples I see that use limma for diff expression use raw counts, which according to limma's documentation, are then converted to logCPM. Can I use the read count files I already have from ballgown?

Also, please let me know if I've left out important information - I'm still new to this. Thanks!

Answer 1

[ATpoint]

Yes, limma is indeed a great choice for differential gene expression. I would stay away from ballgown though, it has never been developed for differential gene expression analysis and fpk is a poor normalization method that does not play well with limma. I assume you have bam files, do you? Use featureCounts together with the GTF annotation file for your organism, make a count matrix with that tool and feed that into limma. limma has an extensive manual you find it at Bioconductor, same goes for featureCounts as part of the Rsubread package.

[khf0008]

That all makes sense to me. Thank you!