Distance summary

Distance Summary computes a genetic distance between every pair of selected sequences and summarises those distances by the taxonomic level at which each pair is related: within a species, within a genus, within a subfamily, within a family or within an order. Use it to see how much divergence exists inside the taxa in a recordset and how that compares across ranks.

Inputs

Records of the chosen marker that pass the shared filters (length, the fixed 10% ambiguity threshold, and a stored sequence). The convert stage writes each record to FASTA with a header of record ID, phylum, class, order, family, subfamily, genus and species; a missing rank is written as an empty field and treated as absent.

The distance is computed column by column across the two sequences, so the sequences must be aligned to a common length. Choose MAFFT under "Align Sequences" unless the stored sequences are already aligned. With None, unaligned sequences of different lengths are still compared position by position up to the shorter length, and the distances are meaningless.

Parameters

Because the report lists parameters as submitted, a report can say "Kimura Two-Parameter" above a table of p-distances. Read every distance on this page as an uncorrected p-distance.

Method

1. Sanitise. Each sequence is upper-cased, and every character other than A, C, G or T, including ambiguity codes, becomes a gap.
2. Distance. For each pair of records, the job counts the positions where neither sequence has a gap (the comparable positions) and the positions among those where the bases differ. The distance is 100 × differences ÷ comparable positions, rounded to two decimal places. Skipping any position with a gap in either sequence is pairwise deletion. A pair with no comparable positions gets a distance of 0.0, not a missing value.
3. Rank assignment. A pair is reported only when both records have a species name. The pair is assigned to the first rank, in the order species, genus, subfamily, family, order, at which the two records carry the same name. A rank that is missing on either record is skipped. A pair whose records differ at every rank up to order is not reported at all.
4. Summary. For each rank and each taxon at that rank, the job reports the number of pairs, the number of distinct records involved, and the sum, minimum, maximum and mean of the pair distances.

The values null, none, mycar and an empty string count as a missing rank.

Outputs

• distances.tsv: every reported pair with its rank, taxon, the two process IDs and the distance.
• distances_species.tsv, distances_genus.tsv, distances_subfamily.tsv, distances_family.tsv and distances_order.tsv: the same rows split by rank. A file is written only for ranks that have at least one pair.
• stats.jsonl: the per-taxon summary. Despite the extension, it is a tab-separated table with the columns rank, taxon, count, records, sum, min, max and avg.
• filtered_records.jsonl: the records included.

aligned_sequences.fas appears in the report's output list but is not produced. The aligned FASTA is not part of the package.

The report contains a Pairwise Distance Distribution histogram in 0.5% bins, with within-species pairs drawn over all other pairs (labelled "between species", which here means every pair assigned to genus or above); a Distance Summary by Species table of the twenty species with the most pairs, with record count, pair count and the minimum, mean and maximum distance; and Mean Distance by Genus and Mean Distance by Subfamily bar charts for the twenty taxa with the most pairs, showing the min–max range and the mean.

Limitations and interpretation

• One distance only. There is no model correction. If a K2P or other corrected distance is required, it must be computed outside the platform from distances.tsv or from the records.
• No overlap control. Two sequences that barely overlap in the alignment produce a distance from very few positions, and two that do not overlap at all produce 0.0. Neither case is flagged.
• Rank assignment is by name match. Two records that share a genus name but belong to different families (a homonym) are grouped as the same genus. Pairs involving a record without a species name are never reported, even for the higher ranks.
• Within-order ceiling. Pairs from different orders are dropped, so the histogram's "between species" set never includes the most distant pairs in a mixed recordset.
• Cost. Every pair is computed in Python, so run time grows with the square of the record count.
• The result does not record the MAFFT version or any alignment settings beyond "automatic".

Related

• The methods and their shared pipeline
• Barcode gap
• Running an analysis
• Known limitations