Output files

Output files

File	Description
all_dist.tsv	Square distance matrix, tab-separated. The first cell is #query, as in mash dist -t.
all_dist.phylip	Same matrix in relaxed phylip format (--phylip). Spaces in names are replaced by _.
sample_stats.tsv	One row per sample (see below).
all.msh	All the sketches in a single file. Can be reused with mash dist or mash screen.
sketches/	One sketch per sample (.msh) and how it was made (.json), reused on the next run.
tree/all_dist_hc.nwk	Hierarchical clustering tree (UPGMA by default).
tree/all_dist_nj.nwk	Neighbour joining tree (--nj).
tree/all_dist_me.nwk	Balanced minimum evolution tree (--me).
tree/all_dist_PCoA.html	Interactive PCoA plot (--pcoa). Self-contained, works offline.
tree/all_dist_PCoA.tsv	PCoA coordinates of each sample on the first 3 axes (--pcoa).
genome_comparator.log	Log of the run.

sample_stats.tsv

Column	Assemblies	Reads
sample	Sample name	Sample name
type	fasta	fastq
files	Number of files	Number of files (2 for paired-end)
length	Assembly size (bp)	Genome size estimated by Mash (bp)
sequences	Number of contigs	Number of reads
est_coverage		Coverage estimated by Mash
status	ok or failed	ok or failed

Trees

• Tip labels are always single-quoted so sample names can contain any character. Single quotes in names are doubled ('it''s'), as the Newick standard requires.
• Branch lengths are Mash distances. In the _hc tree, branch lengths are half the merge heights, so the distance between two tips along the tree matches the clustering distance.
• NJ and minimum evolution trees are unrooted. Negative branch lengths from NJ are set to 0.
• With --bootstrap, support values (0–100) are written as internal node labels (see Bootstrap support).

View the trees in FigTree, iTOL or Dendroscope.

PCoA

The principal coordinates analysis (classical multidimensional scaling) is the equivalent of a PCA for a distance matrix. The axis titles give the percentage of variance explained. Hovering over a point shows the sample name and the metadata columns (--metadata). See How it works to interpret it.

With --color-by, each category gets its own colour and marker shape, so groups can be told apart without relying on colour alone:

• Colours come from the colourblind-friendly Okabe-Ito palette, in a fixed order. They were checked for all common colour vision deficiencies with every pair of colours side by side.
• Categories are assigned in alphabetical order, so a category keeps its colour and shape between runs.
• 6 colours × 7 shapes give 42 unique combinations. If there are more categories, the least frequent ones are grouped into "Other" (grey).
• Samples missing from the metadata file are shown as "Unknown" (grey open circles).