Resolving the food web to who eats whom
The eukaryotic marker was read as a coarse soil food web - fungi at the base, then protists and the small animals above them - across the cores that survived filtering. It was left deliberately coarse, because feeding mode cannot be read off a broad marker above the genus level, and the site says so wherever the food web appears. That coarseness is exactly what a follow-up could lift.
The protist compartment - the second-largest eukaryotic group after the fungi, and present in every core - was dominated by a lineage that is trophically mixed: some of its members graze bacteria, others do not. Published trait databases now assign many of those genera to a feeding mode directly. Applying them would turn "protists" into an estimate of grazing pressure per core - a number with a mechanism behind it, instead of a category on a bar chart.
And it opens a test the paired markers were built for and this study never ran: the bacterial and eukaryotic reads came from the same soil, so the grazing pressure inferred from the eukaryotes can be set directly against the bacterial community measured beside it. Whether bacterivore load tracks bacterial composition or diversity across the plant patches is a genuine predator–prey question, answerable - carefully, at this sample size - from sequences already in hand.
What it cannot become is a large-sample network. Ten to thirteen cores are too few to infer a co-occurrence web whose edges would hold up. That leaves the trophic annotation and the single coupling test, both reported with the sample size in plain view.
Describing this soil alongside published grassland
A single patch cannot be shown to be unusual, because there is nothing beside it to read it against. What the assembled comparison tables allow instead is modest and worth doing: to place this soil's genus-level composition alongside published grasslands and describe where it falls. The aim is a picture of the range and this ground's position in it, not a verdict on the ground - and certainly not on the farm, which these cores do not sample. Note the mismatch to state plainly throughout: the published sets are mostly managed swards, and this is a patch of volunteer vegetation.
The arithmetic already pointed that way. Tested against the published groups, this soil's ammonia-oxidizer share did not separate out (p = ), and neither did its dominant bacterial phylum (p = ). So the comparison would sit this ground somewhere inside the range of working grassland.
One confound would need watching throughout. The public datasets do not all use the same primers, and a marker difference can look exactly like a biological one, so the comparison is only worth as much as its handling of that.
What both of them stop short of
Either one would give a fuller description of this soil, and neither would show that the way the farm is run produced what is in it - nothing can, from cores taken in one weedy patch with no comparison. That takes measurements this study does not have, and they are set out under ground truth.
There is a third direction, and it is cheaper than either of these: sample the same five plants in two or three more patches on the same farm, writing down where each core came from. The plant result was clean precisely because everything came from one patch - but that also means it has only been shown once, in one place. Repeating it is the difference between a pattern and a fact, and it needs no public dataset at all.