Question regarding paralogs

[jessydit]

Hello!

I'm running Panaroo on a bunch of bacterial genomes (different strains of the same species) and I noticed a few things in the output that I cannot explain based on what I read in the documentation. I hope you can help me!

Here is the command I ran:
panaroo -i Seqs/*.gff3 -o Output_aligned/ --clean-mode moderate -a core --aligner mafft --core_threshold 1.00 --remove-invalid-genes --refind-mode off -t 16

As you can see, I did not use "merge_paralogs", so I expected that multi-copy genes would be split into different gene clusters. However, when I inspected the aligned core gene sequences, I noticed that several of these alignments contained more sequences than the number of genomes I used. In the most extreme case, the alignment file contained 90 sequences although I only included 29 genomes in the analysis. I re-ran the same analysis with --aligner none to obtain also the unaligned sequences for this gene cluster. In this unaligned sequence file, I could see that some genomes have one long sequence in this cluster (9800 bp), while other genomes have 5 or 6 short sequences (180-500 bp long).

I have three questions regarding this observation:

1. Is this behaviour intended or is there an error in the gene clustering step? Shouldn't multiple gene copies from the same genome be split into different clusters?
2. I'm astonished by the length difference between the sequences in this gene cluster. I did not modify the default parameters, but I thought that big differences in size would be avoided by the len_dif_percent parameter (default: 0.98)?
3. Despite the multiple gene copies from the same genome in some gene clusters, the global core_gene_alignment.aln file contains 29 sequences, i.e. one for each analysed genomes. That makes me wonder how Panaroo dealt with the multiple gene copies in the single-gene alignments to produce the global alignment? Is the global alignment produced independently from the single-gene alignments or is there an additional step to exclude these multiple copies?

Thanks a lot for looking into this so that I can understand if I need to adjust certain parameters.
Cheers!

[gtonkinhill]

Hi,

This is a result of the way Panaroo attempts to deal with fragmented genes. If Panaroo identifies a long gene in one genome and fragments of this gene in other genomes, then they will be clustered together, leading to the behaviour you describe.

In general, this helps to deal with genes that have been fragmented due to annotation errors and/or misassemblies. In some cases, the long genes are themselves errors. Panaroo attempts to determine this by requiring that the long genes are supported by a specified fraction of the genomes provided (1% by default). This can be adjusted using the --length_outlier_support_proportion parameter. If you increase this value, Panaroo should start to split apart the clusters you’ve observed.

I hope this helps, and please feel free to follow up with additional questions.