funannotate update - antiSMASH: Multiple CDS features have the same location ERROR

[vikll]

Hi,
I am trying to work with RNAseq data in the funannotate pipeline. After running funannotate update there are multiple CDS features with the same coordinates in my output file. When running antiSMASH (online, remote, or local) I am getting the Error:

ERROR 04/07 11:36:32 Multiple CDS features have the same location: join{545164:545343, 544620:545113, 544264:544573}

Command
Predict
funannotate predict -i masked.fasta -o predicted -s "Species" --isolate C1 --cpus 30 --busco_db ascomycota_odb10 --transcript_evidence trained/training/trinity.fasta --rna_bam trained/training/trinity.alignments.bam --transcript_alignments trained/training/funannotate_train.transcripts.gff3

Update
funannotate update -i predicted --cpus 30 --left RNA_S1/R1824_1.fq.gz --right RNA_S1/R1824_2.fq.gz --jaccard_clip --species "Species" --isolate C1

antiSMASH
antismash --taxon fungi --fullhmmer --cassis --clusterhmmer --tigrfam --asf --cc-mibig --cb-general --cb-subclusters --cb-knownclusters --pfam2go --rre --smcog-trees --tfbs --output-dir antiSMASH --logfile antismash.log --genefinding-tool none predicted/update_results/Species_C1.gbk

Funannotate Version
Version 1.8.15

Is there any way to overcome this and solve this issue rather than manually deleting all the dublicate CDS'?

Thanks very much in advance

[#876]

[igwill]

I also had this issue, in my case, it was rare enough (~ 20 genes) for a bit of crude manual checking and removing just for the sake of the antismash step, but keeping the main funannotate gbk/annotation with those duplications.

For me, I got this when one gene locus had multiple mRNA/exon splicing isoforms along with different UTR annotations that ended up leading to the same CDS product.

If that's the same for you, you could search the cds.fa for duplicate sequences, do some sorting/filtering of the seq names to see if they are indeed at the same locus (eg. FUN_00001-T1, FUN_00001-T2 and FUN_00001-T3 all having the exact same CDS sequence) and the remove all but one CDS from the gff.
Then, use antismash with the scaffold.fa as input and the --genefinding-gff3 option pointing to your edited un-duplicated gff.

It ain't pretty and doesn't resolve any underlying issues about if the annotations are good, how to select which copy, what problems this might indicate. But it doesn't take long and does let you proceed if that's okay for your purposes. I did notice that my final annotation still got predicted antismash clusters, but not backbone genes and the notes column in the annotation.txt didn't have any SMCOGs. So again, this might not give you an output worthy of a final assembly, but you can get something to mess around with.

[vikll]

Hi,

Yes, I solved this problem in a similar way now. In my case this occured also when one gene locus had multiple isoforms of a gene with different UTR annotations but I also had some cases where they were completely identical.

I edited the tbl file, run funannotate fix again (so that the gbk will be adjusted) and run antismash with the adjusted gbk file. I did this as I read in another issue that if something must be changed it is recommended to change the tbl file (#296) and editing the gff and afterwards converting it into a tbl with funannotate util (gff2tbl) did not work for me.

I am not sure though if I am going to run funannotate annotate with the adjusted data or with the data before the adjustment.

Thanks for your comment and recommendations