Notebook

NB17 — Cross-cutting synthesis + clinical-translation roadmap (Pillar 5 closure capstone)¶

Project: ibd_phage_targeting — Pillar 5 closure capstone Depends on: NB02 (ecotypes) + NB05 (Tier-A scoring) + NB06 (modules) + NB09c (BA-coupling) + NB10a (mediation) + NB12 (phage availability) + NB15 (per-patient cocktails) + NB16 (longitudinal + dosing rule)

Purpose¶

NB17 is the project's cross-cutting synthesis. Three deliverables consolidate Pillar 1-5 into a single clinical-translation package:

Per-patient master table — one row per UC Davis CD patient with all per-pillar attributes, design category, recommended cocktail strategy, and longitudinal status. The single artifact a clinical collaborator would actually use.
Target decision matrix — 6 actionable Tier-A × 5 attributes (NB05 score, ecotype membership, BA-coupling cost, mediation mode, phage tier) → final cocktail-design priority class.
Clinical-translation roadmap + final verdict — Immediate / Near / Mid / Long term timeline with concrete milestones; final per-pillar verdict + 24 Novel Contributions one-line index.

Per plan v1.9 no raw reads — all inputs are precomputed mart artifacts.

In [1]:

# See run_nb17.py for full source.

§1. Load upstream per-pillar artifacts¶

In [2]:

# NB02 + NB05 + NB12 + NB15 + NB16 outputs

## §1. Load upstream per-pillar artifacts

  NB15 patient profile: 23 patients × 27 attributes
  NB15 cocktail draft (long): 74 patient×target rows
  NB12 phage matrix: 71 species × 8 attributes
  NB16 patient 6967 longitudinal: 6 Tier-A rows × 2 visits
  NB16 verdict: jaccard 0.6; M.gnavus FC 13.96×; 1112 reseq ρ 1.0

§2. Per-patient master table¶

In [3]:

# 23 patients × full per-pillar attributes + design category + cocktail strategy + longitudinal status

## §2. Per-patient master table

  Per-patient master table written: /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/data/nb17_patient_master_table.tsv
  Design-category distribution: {'C_quiescent': 12, 'A_active_many_targets': 8, 'B_active_few_targets': 2, 'D_mixed_longitudinal': 1}
  Cocktail-strategy distribution: {'reserve_for_flare': 12, 'hybrid_3strategy_E1_no_ecoli': 4, 'limited_E0_priority_targets_GAP': 4, 'hybrid_3strategy_E1_full': 1, 'focused_E3_with_ecoli': 1, 'state_dependent_dosing': 1}
  Patients with concrete phage cocktail draft: 14/23

§3. Target decision matrix¶

In [4]:

# 6 actionable Tier-A × 5 attributes → final priority class

## §3. Target decision matrix

  Target decision matrix written: /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/data/nb17_target_decision_matrix.tsv
  Decision matrix:
     species  nb05_score                phage_tier                         ba_coupling_cost                                                                                     priority_class
H. hathewayi         4.0                       GAP                                     none                                                  Tier-1 phage GAP — INPHARED/IMG-VR query priority
   M. gnavus         3.8            temperate-only                                     none                       Tier-1 limited — lytic-locked engineering OR biochemical glucorhamnan target
     E. coli         3.6            clinical-trial                                     none              Tier-1 phage with strain-resolution — NB13 5-phage cocktail; AIEC diagnostic required
    E. lenta         3.3  lytic-literature (PMBT5)     moderate (partial 7α-dehydroxylator)                                                            Tier-2 phage — PMBT5 with BA monitoring
  F. plautii         3.3                       GAP HIGHEST (active 7α-dehydroxylator; F420) Tier-2 deprioritize — triple penalty (GAP + highest BA-cost + E1-only); UDCA/BA-binding co-therapy
  E. bolteae         2.8 lytic-literature (PMBT24)      moderate (active 7α-dehydroxylator)                                                           Tier-2 phage — PMBT24 with BA monitoring

§4. Final per-pillar verdict + Novel Contributions index + clinical-translation roadmap¶

In [5]:

# Pillar 1-5 verdicts + 24 NC index + 4-phase clinical roadmap

## §4. Final per-pillar verdict + Novel Contributions

  Final verdict written: /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/data/nb17_final_verdict.json
  Final status: Pillars 1-5 substantially closed; NB17 cross-cutting synthesis complete
  31 notebooks; 24 Novel Contributions; 6 actionable Tier-A; 14/23 patients with concrete cocktails

§5. Synthesis figure¶

In [6]:

# 3-panel: target decision matrix + per-patient design map + clinical-translation roadmap

## §5. Synthesis figure

  Figure written: /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/figures/NB17_synthesis.png

No description has been provided for this image

§6. Synthesis interpretation¶

Headline: Pillar 1–5 substantially closed; 14 of 23 UC Davis CD patients with concrete cocktail drafts; state-dependent dosing rule operationalized; clinical-translation roadmap defined¶

Per-patient master table — 4 design categories × 7 cocktail strategies¶

The 23 UC Davis CD patients distribute across 4 design categories (combining ecotype + calprotectin + longitudinal status):

Category	n	Description	Cocktail strategy
A — Active disease + many targets (calp ≥ 250 + ≥ 4 actionable Tier-A)	8	High disease burden + multiple intervention points	Hybrid 3-strategy cocktail (NB13 5-phage E. coli if AIEC+; PMBT24; PMBT5; alternatives for H. hathewayi + F. plautii + M. gnavus)
B — Active disease + few targets (calp ≥ 250 + < 4 actionable)	2	High disease burden + limited targets	Limited cocktail; consider non-phage strategies
C — Quiescent (calp < 250 or unmeasured)	12	Low disease activity	Reserve cocktail for flares; calp + qPCR monitoring
D — Mixed-ecotype longitudinal (patient 6967)	1	E1 ↔ E3 drift across visits	State-dependent dosing per NB16 §5 workflow

Cocktail strategies break down further by ecotype + AIEC carrier status:

Strategy	n	Components
Reserve for flare (C category)	12	No active cocktail; calp monitor + M. gnavus qPCR
E1 hybrid 3-strategy (no E. coli)	4	PMBT24 + PMBT5 + alternatives for H. hathewayi + F. plautii + M. gnavus + UDCA/BA-binding
E0 limited (priority targets are GAP)	4	Limited cocktail; consider non-phage strategies
E1 hybrid 3-strategy (full, with E. coli AIEC)	1	NB13 5-phage E. coli + PMBT24 + PMBT5 + alternatives for H. hathewayi + F. plautii + M. gnavus + UDCA
E3 focused (with E. coli)	1	NB13 5-phage E. coli + PMBT5 + alternatives
State-dependent dosing	1	Patient 6967: cocktail rebalances on E1↔E3 ecotype shifts (drop F. plautii on E1→E3; add E. coli AIEC on E3→E1)

Patients with concrete phage cocktail drafts: 14 of 23 (61 %).

Target decision matrix — 6 actionable Tier-A × 5 attributes → priority class¶

The decision matrix integrates Pillar 1-4 evidence per actionable target:

Species	NB05	Ecotype	BA cost	Phage tier	UCD prev.	Final priority class
H. hathewayi	4.0	E1+E3	none	GAP	83 %	Tier-1 phage GAP — INPHARED/IMG-VR external DB query priority
M. gnavus	3.8	E1+E3	none	temperate-only	91 %	Tier-1 limited — lytic-locked engineering OR biochemical glucorhamnan target
E. coli	3.6	E3-specific	none	clinical-trial	35 %	Tier-1 phage with strain-resolution — NB13 5-phage cocktail; AIEC diagnostic required
E. lenta	3.3	E1+E3 (universal Tier-1)	moderate	lytic-literature (PMBT5)	70 %	Tier-2 phage — PMBT5 with BA monitoring
F. plautii	3.3	E1-specific	HIGHEST	GAP	78 %	Tier-2 deprioritize — triple penalty; UDCA/BA-binding co-therapy
E. bolteae	2.8	E1+E3	moderate	lytic-literature (PMBT24)	83 %	Tier-2 phage — PMBT24 with BA monitoring

The matrix exposes the structural shape of the clinical-translation problem:

The 2 highest-NB05-scored species (H. hathewayi 4.0, F. plautii 3.3) are both Pillar-4 GAP — INPHARED + IMG/VR external DB queries are the highest-priority external-data extension.
The species with the most complete phage-therapy precedent (E. coli, clinical-trial via EcoActive) is also the rarest in UC Davis (35 % carriage).
F. plautii has triple penalty (GAP + highest BA-cost + E1-only) → deprioritize from cocktail despite being NB05-actionable.
M. gnavus is near-universal (91 %) but temperate-only — biggest unmet need from a coverage standpoint.

Clinical-translation roadmap¶

Concrete milestones organized by feasibility and timeline:

Immediate (current cohort) — already delivered by this project:

Per-patient cocktail drafts for 23 UC Davis patients (NB15)
F. plautii BA-coupling-cost annotation per actionable target
4-category patient stratification (Active+many, Active+few, Quiescent, Mixed)
5 state-dependent dosing rules + clinical workflow (NB16)

Near-term (6–12 months) — feasible external-data extensions:

INPHARED + IMG/VR external DB queries for H. hathewayi / F. plautii / M. gnavus phages
AIEC strain-resolution diagnostic for the 8 / 23 E. coli-positive patients
M. gnavus qPCR validation as cheap ecotype-state proxy

Mid-term (12–24 months) — additional cohort/assay generation:

Targeted qPCR ecotype panel (4-6 species)
Per-patient bile-acid panel for F. plautii BA-cost monitoring
Multi-cohort serology meta-analysis (firms up H3e PARTIAL)
Expanded longitudinal sampling beyond patient 6967

Long-term (24+ months) — clinical pilot territory:

Clinical pilot of hybrid 3-strategy cocktails (per-ecotype + per-patient + state-dependent)
Lytic-locked phage engineering for M. gnavus
GAG-degrading enzyme inhibitor screening for H. hathewayi

Final thesis (one sentence)¶

Crohn's disease at the gut-microbiome level is a single principal-direction phenomenon (NB07d CC1 r=0.96) within which 6 actionable Tier-A pathobionts and 2 cross-corroborated mechanism narratives (iron-acquisition + bile-acid 7α-dehydroxylation) define a state-dependent, hybrid-cocktail design framework with concrete per-patient drafts for 14 of 23 UC Davis CD patients.

Per-pillar final verdict¶

Pillar	Verdict	Notebooks	Key finding
1 — Patient stratification	CLOSED	NB00, NB01, NB01b, NB02, NB03	4 reproducible IBD ecotypes; HMP2 external-replication χ² p=0.016; UC Davis distributes E0 27%/E1 42%/E3 31% (χ² p=0.019)
2 — Pathobiont identification	CLOSED	NB04(b–h), NB05, NB06	Within-ecotype × within-substudy meta yields 51 E1 + 40 E3 Tier-A; 6 actionable Tier-A core; HMP2 88.2% E1 sign-concordance
3 — Functional drivers	CLOSED	NB07a/b/v18/c/d, NB08a, NB09a/b/c/d, NB10a, NB11	8 H3 verdicts (5+1+2+1); 2 cross-corroborated 6-line narratives; CC1 r=0.96 collapses all narratives to single joint axis
4 — Phage targetability	CLOSED	NB12, NB13, NB14	3-layer phage-evidence convergence; 5-phage E. coli AIEC cocktail at 95% strain coverage; H. hathewayi + F. plautii GAP confirmed all 3 layers
5 — Per-patient cocktails	SUBSTANTIALLY CLOSED	NB15, NB16, NB17	14/23 patients with concrete cocktails; pure phage infeasible for E1; F. plautii BA-cost dominant E1 constraint; patient 6967 E1→E3 drift validates state-dependent dosing rule (M. gnavus 14× expansion); 5 dosing rules + clinical workflow

24 Novel Contributions — one-line index¶

See data/nb17_final_verdict.json field novel_contributions_index for the full list. Categories:

Methodology (general portable patterns): #1 cross-method ARI K-selection, #2 OvR-AUC vs per-patient gap, #6 cMD substudy-nesting, #7 feature leakage, #8 within-ecotype × within-substudy design, #9 adversarial review, #10 LOSO ARI, #12 ontology > regex, #15 pool ≠ flux, #19 cohort-batch in m/z metabolomics, #21 multi-omics joint factor collapse, #22 3-layer phage convergence
Project-specific findings: #3 Kaiju↔MetaPhlAn3 asymmetry, #4 synonymy layer, #5 4-ecotype IBD framework, #11 operational-Tier-A despite framework-variance, #13 module-level metabolic-coupling, #14 5-line iron narrative, #16 BA-coupling cost, #17 two 6-line narratives, #18 species-vs-strain mediation axis, #20 9 strict + 3 theme-level cross-cohort metabolomics replications, #23 hybrid cocktail necessity, #24 ecotype-drift state-dependent dosing rule + qPCR proxy

Out-of-scope (flagged for follow-up)¶

The project's central science question is fully answered within plan v1.9 (no raw reads). The following are flagged as the highest-priority external-data extensions for clinical-translation but out of project scope:

INPHARED + IMG/VR external phage DB queries for the 3 gut-anaerobe phage-coverage gaps (H. hathewayi, F. plautii, M. gnavus lytic alternatives) — closes the structural Pillar-4 gap.
Multi-cohort prospective validation of state-dependent dosing rule — the rule is derived from n=1 longitudinal trajectory (patient 6967); needs expanded longitudinal sampling.
Per-patient AIEC strain-resolution diagnostic — the NB13 5-phage cocktail's clinical applicability requires per-patient strain genotyping.
Clinical pilot of hybrid 3-strategy cocktails — long-term goal; requires regulatory pathway not addressed in this project.

Outputs¶

data/nb17_patient_master_table.tsv — 23 patients × full master attributes + design category + cocktail strategy + longitudinal status
data/nb17_target_decision_matrix.tsv — 6 actionable × 5 attributes + final priority class
data/nb17_final_verdict.json — per-pillar verdicts + 24 NC index + clinical-translation roadmap + thesis statement + design/strategy distributions
figures/NB17_synthesis.png — 3-panel: target decision matrix + per-patient design map + clinical-translation timeline

Nb17 Synthesis