Design Experiments From Reported Results

 BGPT Odds of True

The hypothesis is a causal mediation claim: that microbial allele/gene sweeps in rhizosphere drive host allele-frequency trajectories. Microbial allele shifts are empirically real, but host-mediation via 6 pooled annual time points with 12 plots and unenumerated abiotic controls makes this unproven; the estimate reflects moderate plausibility with wide uncertainty.

 Hypothesis Novelty

Longitudinal paired metagenome–host-WGS mediation linking microbial sweeps to host selection is a relatively fresh integrative design, though component methods (allele-frequency tracking, mediation) are established.

 Quick Analysis Plan Copied

 Long Analysis Plan

Strengths of the Proposed Design

The paired time-series structure (12 plots/site, quarterly metagenomes, 6 years) gives 24 microbial time points and 6 host generations — enough resolution to detect genuine sweeps rather than drift. Pooling 50 host plants per plot for annual WGS trades individual-level resolution for plot-level allele-frequency trajectories, which is appropriate for selection-slope estimation. Sequencing candidate isolates with long reads only after metagenomic screening is a cost-efficient two-stage design. The falsification criterion is clearly stated and pre-registrable, which is a scientific strength.

Empirical Anchors From Supplied Evidence

Microbial alleles do sweep under selection pressures: in V. vulnificus, allele 1 dominates clinical isolates (87%) while allele 2 dominates environmental isolates (87%), P < 0.0001, and phase-variation frequencies differ 2–9× between strains Vibrio allele sweep precedent. This establishes that fast microbial allele shifts are empirically real and context-dependent, but these data come from pure-culture in vitro dynamics — rhizosphere dynamics involve hundreds of interacting taxa plus host feedback, so extrapolation is weak.

Critical Weaknesses

(1) Mediation inference under serial dependence. With only 6 host WGS time points per plot, mediation analysis will rely on cross-lagged correlations that are statistically fragile. Finite-sample tests on dependent time series are often liberal (anti-conservative) when standard asymptotic or i.i.d. bootstrap methods are used Finite-sample inference caution — the design should pre-specify block bootstrap or permutation-by-plot inference, not default frequentist mediation. (2) Pooled WGS loses individual-genotype covariance, preventing host genotype–microbe association at the individual level; allele-frequency covariance across only 6 annual points cannot separate host selection from demographic turnover. (3) Abiotic confounders are asserted, not enumerated. "Controlling for abiotic drivers" requires measured covariates (pH, moisture, temperature, nutrient flux) at plot level per quarter — the design omits these. (4) Reverse causality is not addressed: host allele shifts could alter rhizosphere exudates, driving microbial sweeps — the mediation direction is assumed, not tested. A cross-lagged panel or Granger-style test with temporal precedence rules is needed. (5) Multiple-comparison burden: screening "any taxon showing allele-frequency shifts" across hundreds of taxa × hundreds of genes × 24 time points inflates false-positive microbial sweeps; an FDR-controlled sweep-detection pre-filter must be specified. (6) Statistical power is unstated. No simulation or power analysis is given for detecting mediation with 12 plots/site — with plot-level replication of only 12, between-plot variance may swamp mediation signal.

Improvements

Add individual (not pooled) host genotyping for at least a sentinel subset; pre-register a specific microbial-sweep detection threshold (e.g., Δp > 0.1 over ≤4 quarters with FDR q < 0.05); include a plot-level covariate panel measured quarterly; use permutation inference at the plot level to respect dependence; and add a reciprocal cross-lagged model to test reverse causality. Allele-origin assignment methods in admixed populations achieve ~0.99 accuracy in simulation but remain unvalidated in real systems Allele-origin assignment accuracy caveat — analogous caution applies to inferring microbial allele function from metagenomic reads without isolate validation, which the design partially addresses via long-read isolate sequencing.

What Would Disprove the Hypothesis

If, after FDR control and abiotic-covariate adjustment, microbial allele-frequency trajectories show no temporal precedence relative to host allele shifts, or if randomized abiotic manipulations (e.g., watering/nutrient sub-treatments) reproduce host allele shifts without microbial sweeps, mediation fails. Conversely, a single microbial locus whose long-read-validated allele shows temporal precedence plus functional host-fitness effect would strongly support the hypothesis.