Nb09C Cross Feeding Disambiguation
Jupyter notebook from the Metagenome-Prioritized Phage Cocktails for Crohn's Disease and IBD project.
NB09c — Sample-level Paired Metabolomics × Metagenomics Cross-Feeding Disambiguation¶
Project: ibd_phage_targeting — Pillar 3 sixth notebook (second metabolomics)
Depends on: NB07c module-anchor × pathobiont species coupling; HMP2 cMD R-package fetch (paired metabolomics + metagenomics)
Purpose¶
Disambiguate the NB07c cross-feeding hypothesis. NB07c found A. caccae × pathobiont species-level Spearman coupling at +0.39 (E. bolteae), +0.33 (H. hathewayi), +0.31 (M. gnavus), +0.29 (F. plautii) within-IBD-substudy meta — consistent with either (a) butyrogenic cross-feeding (pathobiont substrates → A. caccae butyrate) or (b) shared-environment co-response to the same CD-specific niche.
NB09c uses paired CSM* HMP2 samples (468 samples with both metabolomics and MetaPhlAn3 metagenomics) to test whether a candidate intermediate metabolite (lactate / SCFA / mucin-glycan / bile-acid) shows the cross-feeding pattern: same-sign correlation with both anchor and pathobiont. If anchor and pathobiont don't share metabolite signatures at the sample level, the NB07c coupling is shared-environment, not cross-feeding.
Per plan v1.9 (no raw reads): uses cMD-fetched HMP2 MetaPhlAn3 abundance + mart fact_metabolomics + ref_hmp2_metabolite_annotations.
Tests¶
- Per (species, metabolite) Spearman ρ on 468 paired samples × 8 NB07c species (anchors + Tier-A core pathobionts) × 583 named HMP2 metabolites; BH-FDR per species.
- Cross-feeding-triangle test — strict criteria: anchor ρ × pathobiont ρ same sign AND both |ρ|>0.20 AND both FDR<0.10.
- Curated cross-feeding metabolite panel — direction-of-association profile across 7 themes (SCFAs, lactate, primary tauro-conjugated BAs, secondary BAs, TMA/choline/carnitine, polyamines, tryptophan/indole) for biological interpretation.
# See run_nb09c.py for full source.
§0. Load paired HMP2 metabolomics × metagenomics + species + metabolite map¶
# Pair metab + cMD MetaPhlAn3 abundance by sample-ID code (CSM*); deduplicate
## §0. Load paired HMP2 metabolomics × MetaPhlAn3 metagenomics HMP2 metabolomics rows: 27348904; named metabolite IDs: 592 HMP2 MetaPhlAn3 abundance shape (dedup): (582, 1605) (species × samples) Paired samples (metabolomics ∩ metagenomics): 468 with diagnosis labels: 468
§1. Build paired sample × {NB07c species + named metabolites} matrix¶
# 8 NB07c species × 583 named metabolites; CLR-transform species; log10-intensity metabolites
## §1. Build paired sample × {NB07c species + named metabolites} matrix
NB07c species present in MetaPhlAn3: 8 / 8
A. caccae (anchor): Anaerostipes caccae
B. nordii (anchor): Bacteroides nordii
H. hathewayi (pathobiont): Hungatella hathewayi
F. plautii (pathobiont): Flavonifractor plautii
E. bolteae (pathobiont): Enterocloster bolteae
E. lenta (pathobiont): Eggerthella lenta
M. gnavus (pathobiont): [Ruminococcus] gnavus
E. coli (Tier-A core): Escherichia coli
Species matrix: (468, 8)
Species prevalence in paired samples:
A. caccae (anchor): 0.02 (9/468)
B. nordii (anchor): 0.09 (44/468)
H. hathewayi (pathobiont): 0.24 (113/468)
F. plautii (pathobiont): 0.83 (387/468)
E. bolteae (pathobiont): 0.51 (238/468)
E. lenta (pathobiont): 0.14 (66/468)
M. gnavus (pathobiont): 0.53 (249/468)
E. coli (Tier-A core): 0.50 (236/468)
Metabolomics rows on paired samples × named: 253079
Metabolite matrix: (468, 592)
Metabolites with ≥30% non-NaN coverage: 583 / 592
§2. Per (species, metabolite) Spearman ρ¶
# Spearman ρ on paired samples; BH-FDR per species (separately for each)
## §2. Per (species, metabolite) Spearman ρ on paired samples Computed correlations: 4664 (species × metabolite × paired samples) Top 5 |ρ| metabolites per species (FDR<0.10): A. caccae (anchor) (200 significant): ρ=-0.372 FDR=4.8e-14 docosapentaenoate (C18-neg) ρ=-0.356 FDR=5.3e-13 docosapentaenoate (HILIC-neg) ρ=+0.334 FDR=2.4e-11 adenine (HILIC-neg) ρ=+0.316 FDR=3.9e-10 hydrocinnamate (HILIC-neg) ρ=+0.305 FDR=2.8e-07 urobilin (HILIC-pos) B. nordii (anchor) (254 significant): ρ=+0.421 FDR=9.2e-19 hydrocinnamate (HILIC-neg) ρ=+0.410 FDR=6.2e-18 hydrocinnamate (C18-neg) ρ=-0.375 FDR=9.2e-15 docosapentaenoate (C18-neg) ρ=-0.369 FDR=2e-14 docosapentaenoate (HILIC-neg) ρ=+0.346 FDR=5e-09 cinnamoylglycine (HILIC-neg) H. hathewayi (pathobiont) (10 significant): ρ=-0.222 FDR=0.0018 bilirubin (HILIC-pos) ρ=-0.201 FDR=0.0036 4-hydroxystyrene (C18-neg) ρ=-0.188 FDR=0.0084 phenyllactate (C18-neg) ρ=-0.177 FDR=0.014 spermidine (HILIC-pos) ρ=-0.177 FDR=0.014 glycodeoxycholate (C18-neg) F. plautii (pathobiont) (201 significant): ρ=-0.328 FDR=3.4e-05 C4 carnitine (HILIC-pos) ρ=-0.281 FDR=0.00028 C10 carnitine (HILIC-pos) ρ=-0.280 FDR=4e-07 2-hydroxy-3-methylpentanoate (HILIC-neg) ρ=-0.277 FDR=4e-07 alpha-ketoisovalerate (HILIC-neg) ρ=-0.276 FDR=0.0023 dihydroorotate (HILIC-neg) E. bolteae (pathobiont) (123 significant): ρ=-0.238 FDR=6.5e-05 C16:0 LPC (HILIC-pos) ρ=-0.237 FDR=6.5e-05 uridine (HILIC-neg) ρ=-0.232 FDR=0.0001 NMMA (HILIC-pos) ρ=-0.224 FDR=0.00038 histidinol (HILIC-pos) ρ=-0.220 FDR=0.00032 homoarginine (HILIC-pos) E. lenta (pathobiont) (132 significant): ρ=-0.221 FDR=0.00053 tryptophan (HILIC-pos) ρ=-0.214 FDR=0.00053 linoleoylethanolamide (C8-pos) ρ=-0.212 FDR=0.00053 serine (HILIC-pos) ρ=+0.211 FDR=0.00053 urobilin (C18-neg) ρ=-0.209 FDR=0.00053 methionine (HILIC-pos) M. gnavus (pathobiont) (210 significant): ρ=-0.268 FDR=1.9e-06 3-methyladipate/pimelate (HILIC-neg) ρ=-0.265 FDR=1.9e-06 sebacate (HILIC-neg) ρ=-0.250 FDR=8.2e-06 hydrocinnamate (HILIC-neg) ρ=-0.246 FDR=0.0043 heptanoate (C18-neg) ρ=-0.243 FDR=1.4e-05 hydrocinnamate (C18-neg) E. coli (Tier-A core) (277 significant): ρ=+0.450 FDR=1.8e-21 cadaverine (HILIC-pos) ρ=+0.350 FDR=1.6e-12 2-hydroxy-3-methylpentanoate (HILIC-neg) ρ=+0.314 FDR=7.2e-10 2-hydroxy-3-methylbutyrate (HILIC-neg) ρ=+0.285 FDR=6.3e-05 C16-OH carnitine (HILIC-pos) ρ=+0.274 FDR=0.00033 C58:10 TAG (C8-pos)
§3. Cross-feeding-triangle test¶
# Strict triangles: anchor × pathobiont same-sign + both |ρ|>0.20 + both FDR<0.10
## §3. Cross-feeding-triangle test: same-sign anchor × pathobiont metabolite candidates Cross-feeding-triangle candidates (same-sign, both |ρ|>0.20, both FDR<0.10): 7 Top 30 by min(|ρ_anchor|, |ρ_pathobiont|): B. nordii & F. plautii : caffeine ρ_a=+0.218 ρ_p=+0.230 B. nordii & E. lenta : linoleoylethanolamide ρ_a=-0.225 ρ_p=-0.214 A. caccae & E. lenta : linoleoylethanolamide ρ_a=-0.213 ρ_p=-0.214 A. caccae & E. lenta : urobilin ρ_a=+0.275 ρ_p=+0.211 A. caccae & F. plautii : cholate ρ_a=-0.261 ρ_p=-0.211 B. nordii & F. plautii : cholate ρ_a=-0.288 ρ_p=-0.211 B. nordii & E. lenta : urobilin ρ_a=+0.343 ρ_p=+0.211
§4. Curated cross-feeding metabolite panel — direction-of-association profile¶
# 7 themes × ~30 metabolites: SCFAs, lactate, primary BAs, secondary BAs, TMA/choline/carnitine, polyamines, tryptophan
## §4. Direction-of-association profile for cross-feeding-relevant metabolites
=== SCFAs (fermentation products) ===
butyrate +0.10* +0.13* -0.08 -0.02 -0.04 -0.05 -0.02 +0.04
propionate +0.07 +0.14* -0.14 +0.05 +0.05 -0.12* -0.07 +0.02
acetate -- -- -- -- -- -- -- --
valerate -- -- -- -- -- -- -- --
hexanoate -- -- -- -- -- -- -- --
isovalerate -- -- -- -- -- -- -- --
isobutyrate -- -- -- -- -- -- -- --
A. B. H. F. E. E. M. E.
=== Lactate (cross-feeding intermediate) ===
lactate +0.18* +0.10* +0.03 -0.23* -0.20* -0.07 +0.10 +0.24*
A. B. H. F. E. E. M. E.
=== Bile acids — primary tauro-conjugated (F. plautii substrate pool) ===
taurocholate +0.04 +0.02 -0.01 -0.10 -0.17* -0.03 +0.15* +0.10*
taurochenodeoxycholate +0.06 +0.07 -0.04 -0.08 -0.15* -0.02 +0.11* +0.09
tauroursodeoxycholate -- -- -- -- -- -- -- --
tauro-alpha-muricholate/tauro-beta-muricholate -0.03 -0.15* -0.00 -0.04 -0.04 -0.11* +0.20* +0.06
taurine -0.07 -0.15* -0.03 +0.04 -0.07 -0.10 +0.17* +0.06
A. B. H. F. E. E. M. E.
=== Bile acids — secondary (F. plautii product pool) ===
deoxycholate -0.07 +0.05 -0.12 +0.06 +0.17* +0.03 -0.10* -0.05
lithocholate -0.05 +0.06 -0.12 +0.15* +0.18* +0.07 -0.20* -0.13*
chenodeoxycholate -0.12* -0.13* -0.02 -0.22* -0.02 -0.13* +0.16* +0.22*
hyodeoxycholate/ursodeoxycholate -0.17* -0.14* -0.05 -0.09 +0.01 -0.11* +0.11* +0.10*
ketodeoxycholate -0.12* -0.15* -0.01 -0.19* -0.05 -0.14* +0.14* +0.24*
A. B. H. F. E. E. M. E.
=== TMA / choline / carnitine (H. hathewayi metabolic products) ===
choline +0.04 +0.01 -0.07 -0.16* -0.15* -0.13* +0.11* +0.25*
trimethylamine-N-oxide +0.03 +0.04 +0.02 -0.01 -0.07 +0.06 -0.03 +0.01
betaine +0.06 +0.07 -0.06 -0.20* -0.11 -0.03 +0.06 +0.20*
carnitine -0.04 -0.02 -0.01 -0.18* -0.06 -0.04 +0.04 +0.19*
C16 carnitine -0.13* -0.19* -0.11 +0.01 +0.03 -0.20* +0.11* +0.19*
C18:1 carnitine -0.10* -0.14* -0.11 +0.02 +0.05 -0.17* +0.10* +0.14*
alpha-glycerophosphocholine -0.04 -0.04 +0.00 -0.12* -0.11* -0.10 +0.08 +0.19*
A. B. H. F. E. E. M. E.
=== Polyamines (CD-up pool) ===
putrescine -0.04 -0.09 -0.07 -0.02 -0.15* -0.11* +0.13* +0.19*
N1-acetylspermine -0.12* -0.14* -0.02 +0.08 -0.02 -0.10 +0.06 +0.10
N-acetylputrescine -0.11* -0.16* -0.05 -0.03 -0.07 -0.12* +0.15* +0.15*
spermine -- -- -- -- -- -- -- --
spermidine +0.01 +0.07 -0.18* +0.16* -0.05 -0.05 +0.05 +0.00
A. B. H. F. E. E. M. E.
=== Tryptophan / indole (gut-bacteria-mediated AA metabolism) ===
tryptophan -0.09 -0.10* -0.10 -0.13* -0.10 -0.22* +0.14* +0.25*
indole-3-propionate +0.10 +0.08 +0.08 -0.04 -0.00 -0.05 -0.04 -0.03
indoleacetate -0.01 +0.01 -0.03 -0.07 -0.03 -0.07 +0.02 +0.06
A. B. H. F. E. E. M. E.
§5. NB07c hypothesis: cross-feeding vs shared-environment¶
# Triangle counts per (anchor × pathobiont) pair
## §5. NB07c hypothesis: cross-feeding vs shared-environment
NB07c found A. caccae × pathobiont species-level coupling:
E. bolteae +0.39
H. hathewayi +0.33
M. gnavus +0.31
F. plautii +0.29
E. lenta +0.08
Cross-feeding hypothesis (a): pathobiont substrates -> A. caccae butyrate.
Shared-environment hypothesis (b): both respond to same CD niche.
Disambiguation criteria (NB09c):
- Cross-feeding: A. caccae & pathobiont share same-sign correlation with
a candidate intermediate (lactate, mucin glycans, bile-acid metabolites);
OR butyrate (the A. caccae product) anti-correlates with pathobionts
(negative feedback) — neither perfectly clean.
- Shared-environment: many metabolites correlate with BOTH species in the
same sign without specific cross-feeding metabolites being prominent.
Cross-feeding-triangle candidates per (anchor × pathobiont):
anchor pathobiont n_metabolites
A. caccae (anchor) E. lenta (pathobiont) 2
A. caccae (anchor) F. plautii (pathobiont) 1
B. nordii (anchor) E. lenta (pathobiont) 2
B. nordii (anchor) F. plautii (pathobiont) 2
§6. Verdict + figure¶
# 2-panel: curated panel ρ heatmap (species × metabolite) + top 25 cross-feeding-triangle candidates
## §6. Verdict + figure
{
"date": "2026-04-25",
"plan_version": "v1.9",
"test": "NB09c \u2014 sample-level cross-feeding disambiguation for NB07c hypothesis",
"n_paired_samples": 468,
"n_metabolites_tested": 583,
"n_species_tested": 8,
"n_cross_feeding_triangles_strict": 7,
"butyrate_acaccae_rho": 0.101,
"lactate_acaccae_rho": 0.18,
"narrative": "WEAK cross-feeding-triangle support at strict thresholds. The NB07c species-level coupling does not strongly disambiguate from shared-environment."
}
Wrote /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/figures/NB09c_cross_feeding_disambiguation.png
§7. Interpretation¶
Headline: NB07c cross-feeding hypothesis NOT supported at sample level → reframe as shared-environment co-occurrence; bile-acid 7α-dehydroxylation network is independently identified as a sample-level mechanistic finding.¶
Cross-feeding-triangle test — only 7 strict triangles¶
The strict cross-feeding-triangle criterion (same-sign + |ρ|>0.20 + FDR<0.10 for both anchor and pathobiont) yields only 7 triangles across 8 species × 583 metabolites — far fewer than the cross-feeding hypothesis would predict if pathobionts genuinely share metabolic intermediates with A. caccae / B. nordii. Of the 7:
- B. nordii / F. plautii × caffeine (+0.22 / +0.23) — environmental exposure, not a metabolic intermediate
- B. nordii / E. lenta × linoleoylethanolamide (-0.23 / -0.21) — fatty acid amide, possibly substrate for both
- A. caccae / E. lenta × linoleoylethanolamide (-0.21 / -0.21) — same
- A. caccae / E. lenta × urobilin (+0.28 / +0.21) — both correlate with the bilirubin-reducer commensal signal (NB09a urobilin CD-DOWN)
- A. caccae / F. plautii × cholate (-0.26 / -0.21) — both anti-correlate with primary BA cholate
- B. nordii / F. plautii × cholate (-0.29 / -0.21) — same
- B. nordii / E. lenta × urobilin (+0.34 / +0.21) — same urobilin pattern
None of these 7 candidates is a butyrogenic cross-feeding intermediate. The pattern is instead consistent with shared-niche health-direction co-occurrence: A. caccae, B. nordii, F. plautii, and E. lenta are all (mostly) commensal species that together correlate negatively with primary BAs (cholate) and positively with urobilin — i.e., they co-occur in healthy / normobiotic samples vs CD samples where dysbiosis dominates. This is the shared-environment pattern, not specific cross-feeding.
Curated cross-feeding panel — biologically coherent but does not support cross-feeding¶
| Metabolite | A. caccae | B. nordii | H. hath | F. plautii | E. bolteae | E. lenta | M. gnavus | E. coli |
|---|---|---|---|---|---|---|---|---|
| butyrate | **+0.10*** | **+0.13*** | -0.08 | -0.02 | -0.04 | -0.05 | -0.02 | +0.04 |
| lactate | **+0.18*** | +0.10* | +0.03 | **-0.23*** | **-0.20*** | -0.07 | +0.10 | **+0.24*** |
| choline | +0.04 | +0.01 | -0.07 | -0.16* | -0.15* | -0.13* | +0.11* | **+0.25*** |
| carnitine | -0.04 | -0.02 | -0.01 | -0.18* | -0.06 | -0.04 | +0.04 | **+0.19*** |
| putrescine | -0.04 | -0.09 | -0.07 | -0.02 | -0.15* | -0.11* | +0.13* | **+0.19*** |
| tryptophan | -0.09 | -0.10* | -0.10 | -0.13* | -0.10 | **-0.22*** | +0.14* | **+0.25*** |
Critical observations:
Butyrate × A. caccae = +0.10*** — significant but very weak. Consistent with A. caccae as a butyrate producer at the population level, but the cohort-level correlation is weak. **The HMP2 LC-MS untargeted methods undersample SCFAs (volatile, polar — typically need GC-MS); only butyrate and propionate are detected, not acetate/valerate/hexanoate. The weak butyrate signal is at least partly methodological.
Lactate × A. caccae = +0.18* vs lactate × F. plautii = −0.23* and × E. bolteae = −0.20* — opposite signs. If lactate were a cross-feeding intermediate (pathobiont produces, A. caccae consumes), we would expect either (a) same-sign positive (both increase together if production dominates) or (b) lagged/asymmetric correlation (production-consumption coupling). The opposite-sign pattern is most consistent with A. caccae and pathobionts occupying different metabolic niches at the cohort level — A. caccae favors lactate-rich states; F. plautii and E. bolteae anti-correlate with lactate (consistent with them being non-lactate-utilizing fermenters).
***E. coli* dominates the cohort-level correlation signal: choline +0.25, carnitine +0.19, tryptophan +0.25, putrescine +0.19, cadaverine **+0.45 (the strongest correlation in the entire panel). These match canonical E. coli / Enterobacteriaceae metabolism: lysine decarboxylase → cadaverine; arginine decarboxylase → putrescine; choline + carnitine substrates for trimethylamine pathways. The v1.8 §9 H. hathewayi TMA/choline finding from the cMD pathway-level analysis does NOT strongly replicate at HMP2 sample level — H. hathewayi × choline ρ=−0.07 (NS); E. coli dominates the choline signal. Possible reasons: (a) HMP2 has lower H. hathewayi prevalence (25 % of paired samples) than the cMD studies that drove the v1.8 finding; (b) at the sample level, E. coli's high prevalence (50 %) and abundance variance dominate the cohort-correlation signal; (c) choline is a substrate for both E. coli and H. hathewayi TMA production, but E. coli is the larger contributor in HMP2. This narrows v1.8 §9 — TMA/choline is a combined Enterobacteriaceae + Lachnospiraceae signal, not specifically H. hathewayi.
Bile-acid 7α-dehydroxylation network is an independent strong finding¶
| BA Class | A. caccae | B. nordii | H. hath | F. plautii | E. bolteae | E. lenta | M. gnavus | E. coli |
|---|---|---|---|---|---|---|---|---|
| Tauro-α/β-muricholate (1° tauro) | -0.03 | -0.15* | 0.00 | -0.04 | -0.04 | -0.11* | **+0.20*** | +0.06 |
| Taurine (1° conjugating AA) | -0.07 | -0.15* | -0.03 | +0.04 | -0.07 | -0.10 | **+0.17*** | +0.06 |
| Cholate (1° unconj.) | **-0.13*** | **-0.18*** | -0.10 | **-0.26*** | -0.10 | **-0.13*** | +0.11* | -0.02 |
| Deoxycholate (2° from cholate) | -0.07 | +0.05 | -0.12 | +0.06 | **+0.17*** | +0.03 | -0.10* | -0.05 |
| Lithocholate (2° from CDCA) | -0.05 | +0.06 | -0.12 | **+0.15*** | **+0.18*** | +0.07 | **-0.20*** | **-0.13*** |
| Hyodeoxycholate/UDCA (2°) | **-0.17*** | **-0.14*** | -0.05 | -0.09 | +0.01 | **-0.11*** | **+0.11*** | **+0.10*** |
| Ketodeoxycholate (2° oxidized) | **-0.12*** | **-0.15*** | -0.01 | **-0.19*** | -0.05 | **-0.14*** | **+0.14*** | **+0.24*** |
The pattern is striking and biologically coherent: ***F. plautii, *E. lenta, and E. bolteae — the canonical bile-acid 7α-dehydroxylating bacteria** — show the predicted substrate-product signature: negative correlation with primary tauro-conjugated bile acids (substrates) and positive correlation with secondary unconjugated bile acids (products: deoxycholate, lithocholate). This is the direct sample-level confirmation of the canonical bile-acid 7α-dehydroxylation network operating in HMP2 samples.
By contrast, ***M. gnavus* and E. coli show the OPPOSITE pattern**: positive correlation with primary tauro-BAs, negative correlation with secondary BAs (lithocholate, hyodeoxycholate). Neither species 7α-dehydroxylates; they are part of a different metabolic network. E. coli's positive correlation with ketodeoxycholate (+0.24) and chenodeoxycholate (+0.22) is consistent with E. coli being abundant in CD samples where the bile-acid pool is shifted toward primary forms.
Implication for NB05 F. plautii targeting: a phage cocktail targeting F. plautii would be predicted to further deplete secondary bile acids (lithocholate, deoxycholate) — these are the anti-inflammatory BA forms. F. plautii is one of the few CD-up species that ALSO carries 7α-dehydroxylation activity in this dataset; its depletion may shift the BA pool back toward inflammatory primary tauro-conjugated forms. NB05 Tier-A scoring should incorporate a "bile-acid coupling cost" annotation (parallel to the "metabolic-coupling cost" from NB07c) for F. plautii-targeted cocktails.
NB07c verdict: REFRAMED as shared-environment co-occurrence¶
The cross-feeding hypothesis (a) is NOT supported by sample-level metabolomic-metagenomic correlation evidence. The shared-environment hypothesis (b) is the more parsimonious explanation for A. caccae × pathobiont species-level coupling:
- 7 strict cross-feeding triangles is well below what cross-feeding would predict
- Top triangle metabolites (caffeine, urobilin, cholate) are health-direction biomarkers, not metabolic intermediates
- Lactate signs are opposite between A. caccae (+0.18) and F. plautii / E. bolteae (−0.20 to −0.23)
- Butyrate × A. caccae +0.10 is weak (and partly methodological — LC-MS undersampling)
Reframed implication for Pillar 4 cocktail design: the NB07c "metabolic-coupling-cost" annotation for A. caccae (NB07c §10) is less load-bearing than originally described — depleting pathobionts via phage cocktail is unlikely to substantially reduce A. caccae abundance through substrate loss (because the substrate-product relationship is not detectable at sample level). The cocktail-design narrative simplifies: target pathobionts directly; A. caccae is in a co-occurring commensal cluster but not metabolically coupled.
Bile-acid coupling cost replaces metabolic-coupling-cost as the primary Pillar 4 annotation for the NB05 actionable set:
- ***F. plautii* targeting**: highest-cost — depletes 7α-dehydroxylation activity, shifts BA pool toward primary inflammatory forms.
- ***E. bolteae* / E. lenta targeting**: secondary 7α-dehydroxylation contributors — moderate cost.
- ***H. hathewayi* / M. gnavus / E. coli targeting**: low BA-coupling cost (these species are not in the 7α-dehydroxylation network).
Methodological observations¶
- 583 of 592 named HMP2 metabolites had ≥30 % non-NaN coverage in the 468-sample paired set. The remaining 9 are likely method-specific compounds with high missingness.
- Cross-cohort applicability: the bile-acid 7α-dehydroxylation finding is HMP2-specific in this analysis. Cross-cohort metabolomics replication (NB09b — FRANZOSA_2019 + DAVE_SAMP_METABOLOMICS) is the natural follow-up.
Outputs¶
data/nb09c_species_metabolite_corr.tsv— all 8 species × 583 metabolites Spearman ρ + FDRdata/nb09c_cross_feeding_triangles.tsv— 7 strict cross-feeding-triangle candidatesdata/nb09c_cross_feeding_panel.tsv— curated 7-theme panel direction-of-association profiledata/nb09c_cross_feeding_verdict.json— formal verdictfigures/NB09c_cross_feeding_disambiguation.png— heatmap + triangle scatter