Diatomaceous Earth Added Nothing to Maize Yield in This Trial, and the Soil Bacteria Were Sampled Once per Treatment

The paper's only statistically tested result is a null. Maize plots that received conventional fertilizer plus diatomaceous earth yielded no more than plots that received fertilizer alone. Both out-yielded an unfertilized control, and that contrast drove the significant difference among the three treatments (F = 18.18; P = 0.0011) [1].

The bacterial half is weaker, and the authors say so: they call the study exploratory. Soil bacterial profiles came from one composite sample per treatment at each of two sampling times, plus a single baseline composite taken before the treatments were applied. Those profiles are reported descriptively. The paper states directly that the apparent patterns — including the highest numerical Shannon index in the fertilized-plus-diatomaceous-earth plots at 120 days after sowing — were not tested inferentially [1].

What was actually done

A randomized complete block field experiment, three treatments, five replications: T1, untreated, no fertilizer and no diatomaceous earth; T2, conventional fertilization; T3, conventional fertilization plus diatomaceous earth. Grain yield was measured from the replicated plots. Soil chemistry and 16S rRNA bacterial profiles came from composite analytical profiles [1].

The diatomaceous earth product was applied to the soil at 12 kg per hectare and contained 73.60 percent silicon dioxide by weight according to the product certificate. Soil was sampled from the 0 to 30 cm layer in the inter-row zone. The 16S rRNA V3-V4 region was sequenced on an Illumina workflow, and the raw FASTQ files were deposited in NCBI SRA [1].

Two things follow. First, the amendment rate is small against the mass of soil it entered: 12 kg per hectare spread through the top 30 cm works out to roughly 3 parts per million by mass, assuming a bulk density near 1.3 grams per cubic centimetre — that is my arithmetic, not a measurement from the paper, and the paper reports no soil silicon measurement at all. Second, the abstract lists grain yield, soil chemistry and 16S profiles as the measurements. It reports no plant silicon uptake, no soil silicon availability, and no mechanism connecting the amendment to the microbial community. The route by which diatomaceous earth might plausibly do anything is not examined here.

What the yield result does and does not mean

The yield finding is real within its own terms: five replicated plots per treatment, a single site, a single season, a significant overall F test. The comparison that matters for the product being sold is T2 versus T3, and those two were statistically similar [1].

That is not the same as proving diatomaceous earth has no effect. A trial with five replications at one site in one season has limited power to detect a small additive effect on top of fertilizer, and a modest yield bump could easily fall below the noise. The honest reading is that this trial did not detect a benefit, not that none exists. Anyone citing this paper as evidence that the amendment is useless is overreading it in the opposite direction from the hype.

What the bacterial data can support

Very little about treatment effects. One composite sample per treatment per time point gives no error term. Composite sampling is deliberate averaging: a reasonable way to get a baseline picture of what is in a field, and the wrong tool for asking whether treatments differ, because it removes the plot-to-plot variation you would need to test against.

The phylum list — Acidobacteria, Actinobacteria, Proteobacteria, Firmicutes and Verrucomicrobia — is what you would expect in an acidic tropical agricultural soil. It is a useful descriptive baseline. It is not evidence that the amendment shifted anything.

The phrase "provider-reported species-level classifications" is worth pausing on. It means the taxonomic labels came from the sequencing provider's pipeline. The abstract does not describe independent verification of those assignments, and species-level calls from short 16S amplicon reads are routinely uncertain.

For scale on how much uncertainty sits inside a small 16S survey, look at a recent exploratory acid mine drainage study: seven samples, 663 amplicon sequence variants, no significant alpha-diversity differences among microhabitat types, and 57.91 percent of pooled reads left unclassified at genus. Its most abundant variant matched a database sequence annotated as Ferrovum but was left formally unclassified because bootstrap support was 46 percent, below the 80 percent threshold [9]. Careful authors report that. It is the normal condition of exploratory amplicon work, not a flaw specific to this maize study.

What stronger evidence would look like

A controlled microbiome experiment can support a causal claim when the inoculum is defined, the plant response is replicated and phenotyped, and the organisms are recovered from the treated host. A recent study screened 135 quinoa-associated bacterial isolates as 45 synthetic communities, identified a nine-isolate consortium, and showed it promoted primary root elongation and increased biomass in Arabidopsis thaliana under salinity, with the clearest effects at 120 mM NaCl. RNA sequencing and RT-qPCR pointed to host genes involved in oxidative stress, water and oxygen processes, phenylpropanoid biosynthesis, glutathione metabolism and root epidermis development, and root hair phenotyping showed a shift in where root hairs formed [2].

That is a different system entirely — a model plant, a controlled salt treatment, a defined inoculum — and it does not transfer to field maize. The point is structural: it shows what has to be in place before a microbiome effect can be asserted rather than suggested.

It is also worth lowering the prior that microbial or mineral additives help by default. A trial feeding a Bacillus licheniformis and Bacillus subtilis probiotic to grazing Nellore bulls finished on tropical pasture found reduced pasture intake, reduced total dry matter intake, and reduced average daily gain, with no change in nutrient digestibility and no effect on rumen or cecum morphometrics [5]. The authors concluded the probiotic did not improve performance under those conditions. Additives can fail, and sometimes they cost you.

A note on the framing

This paper is not a mycology study. It sequenced the 16S rRNA V3-V4 region, which targets bacteria and archaea [1, 9]. Fungal communities require different markers and different methods. A recent endophytic fungus study, for comparison, used confrontation assays, cell-free culture filtrate testing, untargeted chemical profiling and molecular docking to characterize Diaporthe sp. PBS30-7 and its antagonism toward two Penicillium species [14]. Nothing in the maize paper speaks to fungi in soil, and it should not be cited as if it does.

Where the evidence runs out

The yield comparison is the only inferential result, and it shows no difference between fertilizer alone and fertilizer plus diatomaceous earth at one site in one season [1]. Whether a smaller effect exists, or whether the amendment matters at a different rate, soil type or rainfall regime, is untested.

The bacterial findings are descriptive profiles from single composite samples. They cannot establish that diatomaceous earth altered the community, cannot establish the direction of any change, and cannot support a mechanism. The Shannon index comparison at 120 days is a number in a table with no variance attached.

The paper reports no soil silicon measurement, no plant silicon uptake, and no functional or activity-based data. The dominant taxa are listed, not linked to any process.

The authors' own framing is the right one to keep: a transparent baseline for future replicated microbiome work in tropical acidic maize soils [1]. That is a modest and accurate description of what this is. The same restraint applies to a related problem — a six-catchment survey of antimicrobial resistance in Nigerian urban streams reported a strong negative correlation between sanitation scores and resistance indicators, and its authors labelled the result an exploratory ecological association rather than causal evidence, pending larger multi-site confirmation [7]. Small-n field studies can generate good hypotheses. They cannot close the question, and this one does not claim to.