wrong gene_id assignment in transcript_models.gtf

[defendant602]

Dear @andrewprzh ,
Thansk for you continue efforts on this great project, but I have another problem regarding with IsoQuant output files. This is my cmd

isoquant.py --reference ref.fa --genedb ref.gtf     --complete_genedb --bam 'multiple bam files' --labels 'multiple labels' --data_type nanopore --stranded forward     --output output --prefix all --threads 32 --sqanti_output --model_construction_strategy sensitive_ont     --report_novel_unspliced false

I find many transcripts were classified as "novel_not_in_catalog" in novel_vs_known.SQANTI-like.tsv, but these transcripts were assigned "novel_gene_***" gene_id attributes in transcript_models.gtf.

transcript242.NC_016195.1.nnic	NC_016195.1	-	370	2	novel_not_in_catalog	KEH30_p11	unassigned_transcript_945	1554	1	-55	-3 -55	-3	extra_intron_novel;terminal_site_match_left_precise	FALSE	True	NA	NA	NA	NA	NA	NA	NA	False	NA	NA	NA NA	NA	NA	NA	NA	5471	7021	NA	0.35	TTAAACACTCAGCCATTTTA	NA	NA	NA	NA	NA	NA	NA	NA	NA
transcript248.NC_016195.1.nnic	NC_016195.1	-	1059	2	novel_not_in_catalog	KEH30_p11	unassigned_transcript_945	1554	1	-50	-3 -50	-3	extra_intron_novel;terminal_site_match_left_precise	FALSE	True	NA	NA	NA	NA	NA	NA	NA	False	NA	NA	NA NA	NA	NA	NA	NA	5471	7021	NA	0.35	TTAAACACTCAGCCATTTTA	NA	NA	NA	NA	NA	NA	NA	NA	NA

NC_016195.1	IsoQuant	gene	5468	7100	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcripts "2"; 
NC_016195.1	IsoQuant	transcript	5468	7100	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript242.NC_016195.1.nnic"; similar_reference_id "unassigned_transcript_945"; alternatives "extra_intron_novel:5783-7045,tes_match_precise:3"; exons "2";
NC_016195.1	IsoQuant	exon	7046	7100	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript242.NC_016195.1.nnic"; exon_number "1"; exon_id "90"; 
NC_016195.1	IsoQuant	exon	5468	5782	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript242.NC_016195.1.nnic"; exon_number "2"; exon_id "91"; 
NC_016195.1	IsoQuant	transcript	5468	7095	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript248.NC_016195.1.nnic"; similar_reference_id "unassigned_transcript_945"; alternatives "extra_intron_novel:6477-7045,tes_match_precise:3"; exons "2";
NC_016195.1	IsoQuant	exon	7046	7095	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript248.NC_016195.1.nnic"; exon_number "1"; exon_id "92"; 
NC_016195.1	IsoQuant	exon	5468	6476	.	-	.	gene_id "novel_gene_NC_016195.1_249"; transcript_id "transcript248.NC_016195.1.nnic"; exon_number "2"; exon_id "93";

I took a look at these transcripts in IGV browser, and they realy were transcripts of gene KEH30_p11, not novel genes.

I am using IsoQuant-3.10.0, why would these wrong gene_id assignments happen? It realy confuses me.

[andrewprzh]

Dear @defendant602

I think the problem is that IsoQuant first classifies novel isoforms using known slice junctions to understand if it belongs to existing gene or a novel one. It may happen that criteria for assigning a novel transcript to a known gene is a bit to strict.

SQANTI-like classification uses a completely different algorithm, which is launched only if --sqanti_output is set.
I think this algorithm is more accurate.

I will revise the entire transcript model construction module and its algorithms soon. There are some known issues, which are not critical, but lead to some discrepancies in the output.

Best
Andrey