I want to do a Differential Exon Usage (DEU) analysis of both mRNA and lncRNA from the same RNAseq samples using DEXSeq. The raw reads were aligned using STAR and GENCODE human v33 as the annotation file. For the DEU, I would like to use a slightly older version (GENCODE v30) for counting reads, since it contains lncRNAs separated into various classes (lincRNA, sense_overlapping, antisense, etc.). Now, I can think of two strategies for the analysis :
Separate the original comprehensive gtf file into subfiles for each category (mRNA and lncRNA) using the "gene type" argument from the 9th column and "exon" from 3rd column of the GTF file, followed by counting of each GTF file using featureCounts (or HTSeq) and subsequent input of the count matrices into R for DEXSeq analysis.
Run featureCounts (or HTSeq) on the full GTF file, read it into R, do DEXSeq analysis, and then separate the results into mRNA or lncRNA by mapping the ENSEMBLID to the "gene type" from the GTf file.
Which one of these approaches would be the correct way to go about it?