StringTie - isoforms not overlapping

[pbarros]

Hi,
I am using StringTie for transcriptome reconstruction and identification of new isoforms.
While I was exploring the "[file]-transcripts.gtf" output file from stringtie I found something that intrigued me... in the example below I show three isoforms resulting from the same gene ("STRG.14686") and the last one was present in the reference annotation. However the start and end coordinates do not match. The first two isoforms end at 394180 bp and 389422 bp, respectively, while the third starts at 396257 bp...

scaffold_96 StringTie transcript 383404 394180 1000 + .gene_id "STRG.14686"; transcript_id "STRG.14686.1"; cov "72.829010"; FPKM "6.506798"; TPM "8.058224";
scaffold_96 StringTie transcript 383404 389422 1000 + .gene_id "STRG.14686"; transcript_id "STRG.14686.2"; cov "61.675678"; FPKM "5.510321"; TPM "6.824155";
scaffold_96 StringTie transcript 396257 398001 1000 + .gene_id "STRG.14686"; transcript_id "STRG.14686.3"; reference_id "scaffold_96.g39603.t1"; ref_gene_id "scaffold_96.g39603"; cov "2963.938721"; FPKM "264.808624"; TPM "327.947357";

Why is StringTie "clustering" these isoforms in the same gene?

[sdriscoll]

first of all let me say that I agree with your interpretation. the third transcript does not overlap the first two and it has given all three the same 'gene_id' value.

the only thing that comes to mind is that the assembly from stringtie, or cufflinks for that matter, is an attempted explanation, and often a simplification, of the alignment data. you may learn more about this by looking at the alignments in an alignment browser such as UCSC or IGV.

while i doubt this, stringtie could be very, very smart and have assembled the third isoform from reads that multimapped between it and the first two transcripts which would imply they are all the same gene but one that is repeated in more than one position.

[pbarros]

thank you for the input sdriscoll ... maybe I was overthinking this

cheers,
pedro