Confused with terminology/output about pirate HOT 1 CLOSED

Hi Filipe,

PIRATE clusters down to 98% (default) using CD-HIT, then performs all-vs-all BLAST/DIAMOND followed by hierarchical MCL clustering at different amino acid similarity cutoffs. The lowest % identity MCL clusters are used as the gene_families/clusters in the output of PIRATE and the highest %id value for that cluster at which all of the isolates are clustered together is reported as the 'threshold' value in the output file (e.g. all 4 sequences in geneA cluster together at 50,60,70,80,90% identity but split into two clusters of 2 sequences at 95% therefore 90% is reported as the threshold value in gene_families.tsv).

As PIRATE uses a much lower default %id value than roary it will almost always report smaller number of genes (gene_families). At higher thresholds (98%) these will be more in line with default roary. PIRATE reports these higher %id clustering as 'alleles' of a gene family.

PIRATE may report some genes with very high copy number per isolate, which may represent related gene families which are poorly resolved by sequence similarity and the rudimentary paralog splitting algorithm used in PIRATE. If this is the case there are a few things I can suggest to improve the resolution of the clustering.

As for the selection of thresholds, it will depend on if you have a clonal or diverse collection of isolates. If you have a reasonably clonal collection I would increase the default % identity threshold values as lower thresholds are providing little useful information and may lead to clustering of closely related genes (Note: you can also increase the CD-HIT clustering cutoff to provide finer scale allelic info for clonal collections). For diverse collections lower thresholds would be more suitable, but comparison to tools like roary, which are designed for relatively clonal collections/species, may not be productive.

All the best,
Sion

from pirate.