lncRNA pipeline using GENCODE files

1

Dear team,

I’m seeking advice regarding an issue in my RNA‑seq analysis pipeline. My objective is to identify differentially expressed lncRNAs. The workflow I’m using includes fastp (trimming), HISAT2 (alignment), StringTie (assembly and quantification), and DESeq2.

For the reference, I used the GENCODE long non‑coding RNA FASTA sequences and the corresponding lncRNA GTF annotation files. However, after running StringTie, the output shows all lncRNAs with zero counts.

In a previous pipeline where I used the full genome FASTA and comprehensive GTF annotation, I was able to detect differentially expressed lncRNAs in this same dataset. This makes me suspect the issue is related to using only the lncRNA‑specific reference files.

Has anyone encountered a similar problem? Do you know how it can be resolved?

Any help would be greatly appreciated.

Kind regards,

2

Hi Ivan,

I’ve replied to your request sent to Galaxy Australia support team. Not sure if you received the reply or not, so I repeat it on the forum. If you use transcripts, you might be better off with tools designed for transcriptome analysis. You can use this tutorial as a guide: Hands-on: De novo transcriptome assembly, annotation, and differential expression analysis / De novo transcriptome assembly, annotation, and differential expression analysis / Transcriptomics

The tutorial is based on Trinity assembly, but you can adapt it to transcriptome.

It is hard to answer your questions without looking at the data, for example, annotation file for lncRNAs. It might not be suitable for read counting.

Kind regards,

Igor

3

What you’re seeing is actually expected in many cases. HISAT2 and StringTie are designed to work with genomic alignments, not a reference containing only lncRNA transcript sequences. If you align only against the lncRNA FASTA, many reads either won’t map correctly or won’t provide the splice information StringTie needs for quantification. That’s why your previous workflow with the complete genome and full GTF worked. I would recommend aligning against the full reference genome with the complete GENCODE annotation, then filtering the resulting expression matrix for transcripts annotated as lncRNAs before running DESeq2. Also double-check that the chromosome names in your BAM and GTF match exactly, since mismatches there can also result in zero counts.