LncRNA isoforms RNA-seq

Question

[Martin B.]

I have filtered potential lncRNAs from a RNA-seq experiment following the usual pipeline for lncRNA identification (filtering by length, coding potential, protein domains, annotation...). For the functional analysis I want to predict interactions of DE lncRNAs with proteins and other RNA species (miRNAs, mRNAs). The problem is that I get several isoforms from cufflinks/cuffcompare in many lncRNA. Which isoform sequences should I maintain for further analysis? The most expressed one for each gene? Or all of them, supposing that each isoform has its own function?

[lieven.sterck]

I would personally retain the longest one (as is also often done for protein coding genes). If you plan to use all isoforms of the other types of genes as well then of course also keep all isoforms for the lncRNA ones.