# Which of the Transcriptome assembly method is best for identifying novel lncRNAs?

I'm working with human samples and I'm trying to identify novel lncRNAs from tumor samples of Prostate cancer. I'm using reference based transcriptome assembly with stringtie. I have seen that stringtie also has De novo mode. But I'm wondering because stringtie is still using the reference genome sequence to guide the transcript assembly, it's just not using the reference annotation.

Is it really a denovo assembly? In this tutorial De novo transcriptome reconstruction with RNA-Seq Check the paragraph De novo transcript reconstruction they mention like below:

"Now that we have mapped our reads to the mouse genome with HISAT, we want to determine transcript structures that are represented by the aligned reads. This is called de novo transcriptome reconstruction"

1. De novo is basically without reference genome right? why stringtie says denovo mode and using genome for alignment?

2. I would like to know whether identifying novel lncRNAs using reference based assembly is best or denovo assembly? [working on human samples for which reference annotation and reference genome is present]. Which is better and why? Which of these methods give more novel lncRNAs?

• I am not an expert in assembly but note that the human reference transcriptome is of very high quality and quite comprehensive. In order to find new reliable transcripts you need high sequencing depth, many replicates and a good read length. I often see people trying to find new transcripts and then ending up with unreliable inconclusive results. Be sure that you 1) try to talk to assembly experts and 2) have the tools and lab capacity to confirm whatever new candidates you find. – ATpoint Aug 26 '20 at 7:44