Improved auto-mapping of imported nodes labeled "Genus sp."

[jimallman]

This request is condensed from a conversation on #opentreeoflife (from 15:57 to 16:54).

It seems we currently map any node labeled "GENUS_sp" (where GENUS could be any taxon of any level: genus, family, etc.) to GENUS. This is a poor heuristic for some trees, esp. in microbial studies, where there are often siblings of the node (i.e., other children of GENUS) clearly identified and mapped.

In these cases, it would be best to map the "GENUS_sp" to an indeterminate species (child) of GENUS. If no other descendants of GENUS are present, map to GENUS as we do now.

@josephwb also suggested a corresponding check in the NexSON validator: "Are taxa sampled to ancestors and descendants?"

[josephwb]

I would say that when siblings of Taxon_sp., or descendants of Taxon (e.g. if Taxon is a family/order/etc.), are present, we leave Taxon_sp. unmapped, as there is no way to map it in a useful way.

[jar398]

I agree with Joseph. The right thing might be to introduce a new taxon,
assuming a description of the taxon could be conjured making it disjoint
from the other already known taxa.

On Thu, Jul 31, 2014 at 3:33 PM, Joseph W. Brown notifications@github.com
wrote:

I would say that when siblings of Taxon_sp., or descendants of Taxon
(e.g. if Taxon is a family/order/etc.), are present, we leave Taxon_sp.
unmapped, as there is no way to map it in a useful way.

—
Reply to this email directly or view it on GitHub
#20 (comment)
.

[mtholder]

The open peyotl issue in this thread is (just summarizing @jimallman 's post above): have the validator flag cases in which a tree has an OTU that is mapped to an ancestral taxon of one of the other taxa in the tree.

This seems very reasonable, but first we'd need either:

1. some new taxomachine (or oti) service, or
2. OTT parsing by peyotl (see Add support for consulting the hierarchy of OTT #54 )

So this issue will have to wait a while.