View Single Post
 04-04-2014, 07:51 AM #1 kajot Member   Location: Germany Join Date: Dec 2013 Posts: 16 DESeq2 - understanding multi-factor DE analysis Hello again dear SeqAnswers, I have another question regarding RNA-seq data processing, this time with DESeq. I had count tables from my experiment generated successfully by DEXSeq counting script. Since I wanted to use DESeq2 this time, without having to re-count all the samples, I used geneCountTable function on my ExonCountSet in R to "compress" all exon bins into genes. 1) I hope this is correct approach to obtaining gene count tables for DESeq ? Now the main question, my study is about gender differences in response to some knock-out. So I have 4 sample groups: Wild-type male Wild-type female Knock-out male Knock-out female I generated all the necessary tables following the DESeq2 vignette: sex strain WTM_1.counts male WT WTM_2.counts male WT WTM_3.counts male WT WTF_1.counts female WT WTF_2.counts female WT WTF_3.counts female WT KOM_1.counts male KO KOM_2.counts male KO KOM_3.counts male KO KOF_1.counts female KO KOF_2.counts female KO KOF_3.counts female KO 2) What is then a proper approach to see differences between gene expression of KO samples that would take into consideration the natural variance between genders in wild-type (like sex-specific genes) ? I know I can manipulate the GLM with the formula, but to be honest I am unsure if I understand the entire meaning behind it. Currently, just for fun, I used formula (design = ~ sex + strain). Now what I got back are 3 tables with fold-changes - Intercept, male vs female, KO vs WT. 3) What is intercept exactly ? Are values in tables male vs. female and KO vs. WT simply pair-wise comparisons, like ones I would get by using formulas: design = ~ sex and design = ~ strain separately ? I know this might take a longer answer so I will patiently wait for somebody's time. I will be extremely thankful for any insight, I am a beginner in RNA-seq , hence those questions might seem trivial for some. All the best!