Mars_Meteorite_EETA79001_Growth_Panel is a panel of monocultures modelled as a community

[realmarcin]

Found curating it for #183. Not a defect in that record's data — a question about whether it belongs in kb/communities/ in its current shape.

What the record says about itself

A defined four-member panel of microorganisms tested for their ability to grow on actual martian regolith…

• taxonomy: E. coli B, Eucapsis sp., Chroococcidiopsis sp. Chr20-20201027-1, Planococcus halocryophilus
• ecological_interactions: 0

And the source never uses "co-culture", "coculture", "consortium" or "community" — not once in 55 KB. The four organisms were each tested on the same substrate, separately.

Why this is worth raising rather than fixing quietly

MicrobialCommunity is the root class, and every other record in the KB is a set of organisms that were together — a co-culture, an enrichment, a consortium, a native community. This one is a comparison of four monocultures sharing a substrate. That is a real and interesting experiment; it is just not a community.

The consequences are small but real:

1. The connectivity audit counts its four taxa as members with no interactions. COMMUNITY_LEVEL connectivity credit is all-or-nothing: one such interaction exempts every taxon #312's work on COMMUNITY_LEVEL credit assumes a record's members are related; here they are related only by having been tested in the same paper.
2. cultivation_setup, which I have just added, describes conditions under which each member was grown individually. I have said so in notes, but a reader scanning the field will read it as the community's.
3. Any KG export will emit four taxa co-occurring in a community node that never co-occurred.

Options

1. Leave it, with the framing explicit — what I have done: notes states it is a panel and that zero interactions is consistent with that. Cheapest, and the record stays discoverable.
2. Model panels as a distinct class or flag — e.g. a boolean or an EcologicalStateEnum-adjacent value meaning "members tested separately". Would let the audit and any export treat it correctly, and there may be other panel-shaped records: worth a survey of records with len(taxonomy) > 1 and ecological_interactions == [].
3. Remove it — it is a legitimate and well-curated record of a real experiment, so this seems wrong.

I have taken option 1 and filed this, because 2 is a schema decision and I did not want to make it inside a curation PR.

A survey worth running before deciding: how many records have multiple taxa and no interactions at all? If this is the only one, option 1 is probably enough. If there are a dozen, the KB has a record class it has not named.

[realmarcin]

Ran the survey the issue asks for. Four records have more than one taxon and zero ecological_interactions:

taxa	record
5	Early_Dental_Biofilm_FiveSpecies
4	Mars_Regolith_Cyanobacteria_Biofertilizer_Panel
4	Mars_Meteorite_EETA79001_Growth_Panel
4	Defined_Multispecies_Enamel_Caries_Model

That splits cleanly, which is useful:

• Two are named "Panel" and are almost certainly the same shape as this issue — members compared rather than combined. Mars_Regolith_Cyanobacteria_Biofertilizer_Panel should be read before assuming, given how often that assumption has failed elsewhere in Enrich records lacking members/growth conditions once full text is accessible #183.
• Two are biofilm/caries models, which almost certainly are communities whose interactions simply have not been curated yet. A five-species dental biofilm is a community by construction; zero interactions there is a curation gap, not a record-shape question.

So the answer to "is this the only one" is no, but it is close to it — one confirmed, one likely, and two that are probably just under-curated.

That makes option 1 (leave it, framing explicit) the proportionate choice, and option 2 (a schema flag) hard to justify for two records. If a third panel appears, the calculus changes.

Worth noting for whoever picks this up: the two biofilm records are a separate, cheaper opportunity — they are communities with no interactions recorded, which is ordinary curation work rather than a modelling decision.

[realmarcin]

Read Mars_Regolith_Cyanobacteria_Biofertilizer_Panel. It is confirmed a panel — so this is two, not one.

Its source (PMID:35865930, 66 KB) uses "co-culture" 0 times, "coculture" 0, "consortium" 0, "community" 0. The record's own description says "a defined four-member panel of photosynthetic microorganisms evaluated for their ability to grow" — Anabaena cylindrica, Nostoc muscorum, Arthrospira platensis, Chlorella vulgaris, each tested on MGS-1 regolith extract.

Same shape as Mars_Meteorite_EETA79001_Growth_Panel, and by the same evidence: separate evaluation, zero interactions, and a source that never describes the four as being together.

Revised tally

record	verdict
Mars_Meteorite_EETA79001_Growth_Panel	panel, confirmed
Mars_Regolith_Cyanobacteria_Biofertilizer_Panel	panel, confirmed
Early_Dental_Biofilm_FiveSpecies	community, interactions uncurated
Defined_Multispecies_Enamel_Caries_Model	community, interactions uncurated

So the split is clean: two panels, two under-curated communities. My earlier comment called the second one "likely" — it is now read rather than assumed, which given how often the assumptions in #543 failed seemed worth doing properly.

Does two change the recommendation?

Marginally. Two records is still thin ground for a schema flag, and option 1 (explicit framing in notes) remains proportionate. But the pattern is no longer a one-off, and both panels are Mars-resource growth assays — if that line of work continues, more will arrive, and a third would make the case.

No cultivation_setup added for this one

Its own growth conditions are not in the cached text. What the cache does carry is the regolith-extract preparation (4:1 water:regolith, autoclaved 121 °C, centrifuged, re-autoclaved) and a Lemna minor bioassay at 23 ± 2 °C under ~74 μmol photons m⁻² s⁻¹ — but Lemna is a duckweed, not one of the four members. Recording those would be #529 at one remove: conditions belonging to a different organism in the same paper.

Left uncurated rather than filled with the nearest available numbers.