pipeline.soillifeatlas.org
Run it
Demo preview · static page · no real compute · kingdom numbers from the ClimGrass decomposition demo

/  soillifeatlas_report.html

run_id

climgrass_20260418_0901

12 ClimGrass grassland samples,
decomposed against atlas v0.1.0.

profile
docker
ion mode
POS
config
iter3a_floor020
runtime
19m 42s

§ 01 · Composition

Kingdom composition
(fc_weighted_bc, primary).

Proportion (%) of each kingdom's distinctiveness in the average sample. Tolerances are the CI assertion bands enforced by the nightly full-demo workflow.

Bacteria

35.73%  ± 2.0 pp

Fungi

29.58%  ± 3.0 pp

Plantae

19.00%  ± 2.0 pp

Archaea

9.13%  ± 3.0 pp

Animalia

3.05%  ± 1.0 pp

Protozoa

3.52%  ± 1.0 pp

assertion

Published values fall within these tolerances; the CI fails otherwise. Numbers above are mean across 12 samples under the primary decomposition method (fold-change weighted Bray-Curtis).

total

100.01%

rounding residual ≤ 0.02

§ 02 · Phylum detail

Phylum-level breakdown.

Each row is the mean proportion across all 12 samples. Phyla within a kingdom are ordered by abundance. Archaeal phyla reflect the ArchLips reference-quality filter (L5).

illustrative demo phylum numbers · kingdom totals match published values

Bacteria

35.73% total · 5 phyla

  • Actinomycetota
    14.82%
  • Pseudomonadota
    11.07%
  • Bacteroidota
    5.94%
  • Bacillota
    2.43%
  • Acidobacteriota
    1.47%

Fungi

29.58% total · 3 phyla

  • Ascomycota
    17.82%
  • Basidiomycota
    10.21%
  • Mucoromycota
    1.55%

Plantae

19.00% total · 1 phyla

  • Streptophyta
    19.00%

Archaea

9.13% total · 2 phyla

  • Thermoproteota
    7.48%
  • Euryarchaeota
    1.65%

Animalia

3.05% total · 2 phyla

  • Nematoda
    1.93%
  • Arthropoda
    1.12%

Protozoa

3.52% total · 2 phyla

  • Cercozoa
    2.17%
  • Amoebozoa
    1.35%

§ 03 · Plausibility

BC distance vs expected
kingdom composition.

bray-curtis

threshold 0.150

0.131

0.15
0.00 0.30

Pass. BC = 0.131 is well below the CI threshold of 0.150. Decomposition converges on a community composition that matches the a priori expected distribution from literature.

per-method agreement

nnls 0.147
std_bc 0.142
enriched_bc 0.135
fc_weighted_bc primary 0.131

All four methods converge within 0.016 BC of each other — confidence that the ecology, not the algorithm, is driving the answer.

§ 04 · Treatment effects

Are the published
climate responses still there?

Mann-Whitney U on per-sample phylum proportions, stratified by treatment. Highlighted rows are the anti-regression assertions enforced in nightly CI.

phylum contrast dir |log fc| p (U-test) status
Actinomycetota Bacteria drought vs control ↑ 0.41 0.003 paper · preserved ✓
Basidiomycota Fungi drought vs control ↓ 0.33 0.011 paper · preserved ✓
Pseudomonadota Bacteria drought vs control ↓ 0.19 0.048 preserved
Ascomycota Fungi warming vs control ↑ 0.14 0.092 n.s.
Nematoda Animalia drought vs control ↓ 0.28 0.024 preserved

Highlighted rows are anti-regression checks · Actinomycetota ↑ drought is the headline assertion from analysis-19.

§ 05 · Diagnostic

Top features per phylum
RIE over-amplification guard.

Confirms the "no single feature dominates" property. If any feature contributes more than 3.1% of a phylum's distinctiveness, the RIE correction is over-amplifying a weak ionizer and the run is flagged.

feature_id m/z rt (min) phylum contribution
ft_00421 752.562 7.83 Actinomycetota 2.91%
ft_02108 688.491 6.12 Ascomycota 2.44%
ft_10772 774.544 9.06 Basidiomycota 2.12%
ft_00019 522.355 4.48 Streptophyta 1.93%
ft_11903 821.612 11.20 Thermoproteota 1.76%

All top features < 3.1% threshold · distributed-signal property holds.

§ 06 · Downloads

Every artefact is
a plain file you can read.

  • composition_fc_weighted_bc.parquet
    142 KB
  • composition_nnls.parquet
    138 KB
  • composition_std_bc.parquet
    140 KB
  • composition_enriched_bc.parquet
    141 KB
  • plausibility.tsv
    0.9 KB
  • treatment_effects.tsv
    3.4 KB
  • diagnostic.tsv
    2.2 KB
  • soillifeatlas_report.html
    1.1 MB
  • pipeline_info/provenance.yaml
    0.6 KB

Downloads disabled on the demo preview · run the pipeline locally to produce real artefacts