Panaroo version: 1.5.2
Command: panaroo -i input_list.txt -o results --clean-mode moderate --remove-invalid-genes --aligner clustal (default refind mode, no --refind-mode strict)
Input: 90 bacterial isolates, Bakta-annotated GFF3s (hybrid assemblies)
Issue: Valid, in-frame, correctly-annotated AMR gene calls are being deleted during clean-mode graph filtering and then recovered via the refinding step, losing their original locus tag in the process (replaced by N_refound_M IDs). When comparing the refound gene sequence to the original (deleted) Bakta annotation it replaced, and to an identical ortholog from another isolate, the alignment is 100% identical over the full coding sequence — but the refound sequence has two extra trailing characters: a * (stop codon) and an X (ambiguous residue), giving it a length of 288 aa vs. 286 aa for the two Bakta-derived sequences.
It doesnt appear to be an annotation-quality problem, as I looked at the translated CDS for one such case and it had no internal stops, 286 aa, 100% identical at the amino acid level to the gene copies elsewhere in the dataset that survived as their original annotation. I also ruled out contig-end fragmentation as the cause.
Question: Is this the expected behavior for genes with variable synteny (e.g. plasmid-borne AMR genes inserting at different flanking contexts across isolates)?
If so:
Is there a recommended way to flag/protect known real gene families? Since refinding doesn't preserve the original locus_tag even when it's recovering the exact same annotation it just deleted, is there a way to retain the original locus_tag / annotation metadata for refound genes that match an existing GFF annotation with 100% identity, rather than replacing it with a refound placeholder? This would materially help downstream traceability for AMR/gene-of-interest analyses.
Another issue that i observed was with fragmented genes (genes with 2 locus tags/sometimes 3 separated by a semicolon). I observed that sometimes 2 fragments are adjacent locus_tags (say ABXT_0096, ABXT_0097) but found on opposite strands and sometimes the 2 fragments were present on 2 different assembly contigs but not the contig ends. Why is the software considering such cases as fragments that belong together?
Appreciate your help with these issues. This is my first time working with Panaroo.
Best,
Priyanka
Panaroo version: 1.5.2
Command: panaroo -i input_list.txt -o results --clean-mode moderate --remove-invalid-genes --aligner clustal (default refind mode, no --refind-mode strict)
Input: 90 bacterial isolates, Bakta-annotated GFF3s (hybrid assemblies)
Issue: Valid, in-frame, correctly-annotated AMR gene calls are being deleted during clean-mode graph filtering and then recovered via the refinding step, losing their original locus tag in the process (replaced by N_refound_M IDs). When comparing the refound gene sequence to the original (deleted) Bakta annotation it replaced, and to an identical ortholog from another isolate, the alignment is 100% identical over the full coding sequence — but the refound sequence has two extra trailing characters: a * (stop codon) and an X (ambiguous residue), giving it a length of 288 aa vs. 286 aa for the two Bakta-derived sequences.
It doesnt appear to be an annotation-quality problem, as I looked at the translated CDS for one such case and it had no internal stops, 286 aa, 100% identical at the amino acid level to the gene copies elsewhere in the dataset that survived as their original annotation. I also ruled out contig-end fragmentation as the cause.
Question: Is this the expected behavior for genes with variable synteny (e.g. plasmid-borne AMR genes inserting at different flanking contexts across isolates)?
If so:
Is there a recommended way to flag/protect known real gene families? Since refinding doesn't preserve the original locus_tag even when it's recovering the exact same annotation it just deleted, is there a way to retain the original locus_tag / annotation metadata for refound genes that match an existing GFF annotation with 100% identity, rather than replacing it with a refound placeholder? This would materially help downstream traceability for AMR/gene-of-interest analyses.
Another issue that i observed was with fragmented genes (genes with 2 locus tags/sometimes 3 separated by a semicolon). I observed that sometimes 2 fragments are adjacent locus_tags (say ABXT_0096, ABXT_0097) but found on opposite strands and sometimes the 2 fragments were present on 2 different assembly contigs but not the contig ends. Why is the software considering such cases as fragments that belong together?
Appreciate your help with these issues. This is my first time working with Panaroo.
Best,
Priyanka