Evidence-ranking matrix for disease-group findings

pre-specified
synthesis
validation
Ranks the strongest diet, mycobiome, and symptom differences across Study_group_new to prioritize candidates for future validation.
Author

IBD Capstone team

Published

May 29, 2026

1 Research question

Which disease-group (Study_group_new) findings are strongest candidates for future validation?

We synthesize tests already run across the capstone pipeline into a single evidence-ranking matrix, ordered by a transparent score combining p-value and effect size.

2 Data and inputs

Domain Tests included Primary inputs
Mycobiome (beta) PERMANOVA × taxonomic level data/processed/{phylum,family,genus,species}.csv
Mycobiome (alpha) Kruskal–Wallis × diversity metric data/intermediate/alpha_long.rds
Diet Kruskal–Wallis × nutrient data/processed/dietary_cleaned.xlsx + mycobiome metadata
Symptoms Kruskal–Wallis × symptom score data/processed/cleaned_characteristics.csv + mycobiome metadata
  • Comparison: Active IBD vs Quiescent vs Non-IBD (Study_group_new).
  • Unit: Sample-level for mycobiome; participant-level for diet and symptoms (diet/symptoms joined to mycobiome participants).
  • Not included here: Continuous symptom–composition associations (Spearman/PERMANOVA on HBI), stratified nutrient–diversity Spearman, or unsupervised PCA clusters — see linked posts below.

Summary: 24 study-group comparisons ranked; 1 with p < 0.05; 1 tagged High priority.

3 Methods

Item Choice
Integration Recompute or aggregate from processed/intermediate files (same sources as individual posts)
Beta diversity PERMANOVA (adonis2), Bray–Curtis, 999 permutations; effect = R²
Alpha, diet, symptoms Kruskal–Wallis; effect = η²[H] (rstatix::kruskal_effsize)
Evidence score −log₁₀(p) × (0.5 + 0.5 × effect_scaled) with R² or η²[H] capped at 1
Validation tier High: p < 0.05 and score ≥ 75th percentile among significant; Medium: p < 0.05; Suggestive: p < 0.10; Low: otherwise
Multiplicity Exploratory ranking; no global FDR across all rows
Show code 
top_validation <- ranked |>
  filter(validation_tier %in% c("High", "Medium")) |>
  slice_head(n = 10)

4 Results

4.1 Full evidence-ranking matrix

Rank Tier Domain Feature Test Effect Statistic p n Evidence score Source post
1 High Symptoms (survey) Harvey-Bradshaw Index Kruskal–Wallis η²[H] = 0.924 10.3150 0.0058 12 2.1544 2026-05-22-symptoms-fungal-composition
2 Suggestive Diet (nutrients) Added sugar (g/day) Kruskal–Wallis η²[H] = 0.423 5.8077 0.0548 12 0.8974 2026-05-29-nutrients-diversity-disease-group
3 Low Mycobiome (alpha diversity) Shannon diversity Kruskal–Wallis η²[H] = 0.116 3.5103 0.1730 16 0.4252 2026-05-19-alpha-diversity-kruskal-wallis
4 Low Diet (nutrients) Fat (g/day) Kruskal–Wallis η²[H] = 0.137 3.2308 0.1990 12 0.3985 2026-05-29-nutrients-diversity-disease-group
5 Low Mycobiome (beta diversity) Species composition PERMANOVA (Bray–Curtis) R² = 0.145 1.1061 0.2660 16 0.3294 2026-05-19-permanova-species-beta-diversity
6 Low Mycobiome (beta diversity) Family composition PERMANOVA (Bray–Curtis) R² = 0.142 1.0773 0.3730 16 0.2446 2026-05-19-permanova-family-beta-diversity
7 Low Mycobiome (beta diversity) Phylum composition PERMANOVA (Bray–Curtis) R² = 0.140 1.0613 0.4120 16 0.2196 2026-05-19-permanova-phylum-beta-diversity
8 Low Diet (nutrients) Vitamin D (mcg/day) Kruskal–Wallis η²[H] = 0.000 2.0000 0.3680 12 0.2171 2026-05-29-nutrients-diversity-disease-group
9 Low Diet (nutrients) Sugar (g/day) Kruskal–Wallis η²[H] = -0.013 1.8846 0.3900 12 0.2045 2026-05-29-nutrients-diversity-disease-group
10 Low Diet (nutrients) Omega-6 (g/day) Kruskal–Wallis η²[H] = -0.013 1.8846 0.3900 12 0.2045 2026-05-29-nutrients-diversity-disease-group
11 Low Diet (nutrients) Calories (kcal/day) Kruskal–Wallis η²[H] = -0.056 1.5000 0.4720 12 0.1630 2026-05-29-nutrients-diversity-disease-group
12 Low Mycobiome (alpha diversity) Chao1 diversity Kruskal–Wallis η²[H] = -0.039 1.4941 0.4740 16 0.1621 2026-05-19-alpha-diversity-kruskal-wallis
13 Low Mycobiome (alpha diversity) Simpson diversity Kruskal–Wallis η²[H] = -0.043 1.4412 0.4860 16 0.1567 2026-05-19-alpha-diversity-kruskal-wallis
14 Low Diet (nutrients) Folate (mcg/day) Kruskal–Wallis η²[H] = -0.064 1.4231 0.4910 12 0.1545 2026-05-29-nutrients-diversity-disease-group
15 Low Diet (nutrients) Artificial sweeteners (mg/day) Kruskal–Wallis η²[H] = -0.090 1.1923 0.5510 12 0.1294 2026-05-29-nutrients-diversity-disease-group
16 Low Symptoms (survey) Fatigue frequency Kruskal–Wallis η²[H] = -0.098 1.1200 0.5710 12 0.1217 2026-05-22-symptoms-fungal-composition
17 Low Symptoms (survey) Abdominal pain Kruskal–Wallis η²[H] = -0.146 0.6875 0.7090 12 0.0747 2026-05-22-symptoms-fungal-composition
18 Low Diet (nutrients) Tryptophan (g/day) Kruskal–Wallis η²[H] = -0.171 0.4615 0.7940 12 0.0501 2026-05-29-nutrients-diversity-disease-group
19 Low Mycobiome (beta diversity) Genus composition PERMANOVA (Bray–Curtis) R² = 0.108 0.7896 0.8520 16 0.0385 2026-05-19-permanova-genus-beta-diversity
20 Low Diet (nutrients) Omega-3 (g/day) Kruskal–Wallis η²[H] = -0.192 0.2692 0.8740 12 0.0292 2026-05-29-nutrients-diversity-disease-group
21 Low Symptoms (survey) Anxiety frequency Kruskal–Wallis η²[H] = -0.204 0.1600 0.9230 12 0.0174 2026-05-22-symptoms-fungal-composition
22 Low Symptoms (survey) Abdominal bloating frequency Kruskal–Wallis η²[H] = -0.204 0.1600 0.9230 12 0.0174 2026-05-22-symptoms-fungal-composition
23 Low Diet (nutrients) Total fiber (g/day) Kruskal–Wallis η²[H] = -0.205 0.1538 0.9260 12 0.0167 2026-05-29-nutrients-diversity-disease-group
24 Low Symptoms (survey) Daily soft stools Kruskal–Wallis η²[H] = -0.222 0.0000 1.0000 12 0.0000 2026-05-22-symptoms-fungal-composition

4.2 Top candidates for future validation

Highest-priority findings (High or Medium tier, top 10 by evidence score)
rank validation_tier domain feature test effect_label p evidence_score
1 High Symptoms (survey) Harvey-Bradshaw Index Kruskal–Wallis η²[H] = 0.924 0.0058 2.1544

Top validation candidates:

  1. [High] Symptoms (survey) — Harvey-Bradshaw Index (η²[H] = 0.924, Kruskal–Wallis, p = 0.0058, score = 2.15)

4.3 By domain (mean evidence score)

Domain Tests p < 0.05 Mean evidence score Top feature
Diet (nutrients) 11 0 0.224 Added sugar (g/day)
Mycobiome (alpha diversity) 3 0 0.248 Shannon diversity
Mycobiome (beta diversity) 4 0 0.208 Species composition
Symptoms (survey) 6 1 0.398 Harvey-Bradshaw Index

5 Figure

Show code 
plot_evidence_heatmap(ranked, top_n = min(20L, nrow(ranked)))
Figure 1: Evidence scores for the top 20 study-group comparisons. Higher scores indicate stronger statistical support for validation.

6 Interpretation

The matrix does not replace individual analysis posts; it prioritizes which disease-group contrasts to carry forward (e.g. pre-registration, larger cohort, or experimental follow-up).

High tier findings combine statistical significance with relatively large R² (composition) or η²[H] (rank-based outcomes). Low tier rows may still be biologically interesting but need replication. Domains with few participants (especially diet and symptoms) should be interpreted cautiously.

7 Reproducibility

  • Helpers: stats/R/evidence_ranking_helpers.R
  • Render: quarto render stats from repository root