#Haplotypes and assignment

#A haplogroup's haplotype

def h as lineage.MtHaplotype init mitoleaf.haplotype($t, "L0a1");
lineage.mtDesignations($h);

The motifs from the root down, with back mutations applied (see Motifs and back mutations). The haplotype's one range is mitoleaf.WHOLE_GENOME, "1-16569", because a motif may name any position.

An observed haplotype usually covers only the control region (HVR-I, or the whole CR), so the ranges disagree:

lineage.mtRangesAgree(mitoleaf.haplotype($t, "L0a1"), $observed);   # false

That is correct. lineage keeps "do the differences match" and "were these sequenced alike" as separate questions, and distance and assign answer only the first.

#Distance

mitoleaf.distance($t, "L0a1", $observed);

lineage.mtDistance over the two difference lists: a designation only one side carries counts 1, so a position both carry with different bases counts 2.

#Assignment

def m as mitoleaf.Match init mitoleaf.assign($t, $observed);
$m.haplogroup;
$m.distance;

The haplogroup with the smallest distance. One depth-first pass in document order carries the accumulated motif state down the tree, so each haplogroup is scored once. Ties go to the first node reached: an ancestor beats an equally distant descendant, and otherwise the earlier node in the file wins.

assign is a nearest-neighbour report, not a haplogroup call; the reasons are in Limitations. Check the result against the tree:

for (def hg in mitoleaf.path($t, $m.haplogroup)) {
    io.printf("%s %v\n", $hg, mitoleaf.motif($t, $hg));
}

#Performance

On the published 6409-haplogroup tree, one desktop core:

Parse the tree once and reuse the Tree. parse builds a flat node list and assembles the tree in one pass; threading the Tree through the recursion deep-copied the node map at every node and took 2m14s on the same file. assign avoids rebuilding each haplotype from the root, which would multiply the work by the tree's depth.