Status
Preliminary — Spearman correlation, db-RDA, and PGLS analyses complete (NB01–NB04). NB05 effect size and spatial validation pending. NB06 figures pending.
Key Findings
- 2,355 significant COG-metal associations (FDR < 0.05) across 9 metals (Cu, Co, Cr, Ni, Zn, Pb, As, Cd, Hg) in 51,748 soil samples. Chromium and Lead drive the strongest signals; transporters (ABC, RND) and biosynthesis genes dominate top hits.
- Copper-specific analysis (116 COGs, FDR < 0.05; 7,566 samples with nearby KBase genomes at 10 km): top positives include cell division and nucleotide transport (BQ, FQ, FK, CF); top negatives include energy production (DI, CE), suggesting energetic trade-offs under copper stress.
- db-RDA: R² = 0.799, p = 0.005 (permutation test, 999 permutations) — metal concentrations explain 80% of variance in community COG profiles after conditioning on batch/project effects.
- Biome-stratified PGLS: soil, marine, and wastewater show distinct metal-COG relationships, indicating environment-specific functional responses rather than a universal resistance programme.
Interpretation
Soil metal concentrations are strongly associated with the functional gene content of co-located microbial communities (db-RDA R²=0.799), and this relationship is not a statistical artefact of batch effects (project accession conditioned out). The direction of COG associations with copper (energetic trade-offs, membrane transport enrichment) is consistent with known copper toxicity mechanisms in bacteria.
However, all associations are observational. Co-contamination by multiple metals is a significant confound: Cr, Cu, Pb, Zn co-vary in many industrial soils. Whether individual COG-metal associations are metal-specific or reflect a generic stress response to multi-metal contamination is unresolved (NB05 partial correlation models).
Effect sizes (Spearman ρ) have not yet been systematically reported across all 2,355 associations; many may be statistically significant but biologically small.
Critical Assessment
R² = 0.799 requires careful interpretation. db-RDA conditions on project accession before fitting the metal concentration predictors. If batch effects (project) account for the majority of community COG variance, the reported R² describes how well metals explain the residual variance, not the total variance. The unconditional R² of metals alone — without first removing project effects — is not reported and may be substantially lower. This must be reported alongside the conditional figure.
2,355 significant associations at FDR < 0.05 in 51,748 samples: With 9 metals × 435 COGs = 3,915 tests, a 60% discovery rate is high. Metals co-contaminate (Cr, Pb, Zn, Cu co-vary in industrial soils), so tests are not independent — Benjamini-Hochberg FDR correction assumes independent tests and will be anti-conservative under positive correlation. True FDR is likely higher than reported. The partial correlation model (NB05 item 3) is the correct remedy.
Copper-specific analysis uses a 10 km proximity criterion to match soil samples to KBase genomes. At 10 km, many “nearby” genomes may not be co-located with the copper measurements. Sensitivity analysis at 5 km and 20 km thresholds is needed before COG-copper associations can be attributed to the measured sites.
Spark-required analyses: NB05 items 1–4 all require re-running from the
kescience_mgnify/kbase_ke_pangenome Spark tables. No local CSV output of the
Spearman ρ values or model residuals is available, so these validations cannot be
completed without Spark access.
Pending Validation
- Effect size audit: ρ distribution across 2,355 associations; flag ρ < 0.05
- Spatial autocorrelation test (Moran’s I on residuals; SEVM if significant)
- Partial correlation model: COG ~ Cr | Cu + Zn + Pb (metal-specific vs. general)
- Mechanistic classification of significant COGs into resistance, stress, membrane, energy, and unknown categories
- Report unconditional db-RDA R² (metals only, without project accession conditioning)
- Copper proximity sensitivity: rerun at 5 km and 20 km thresholds