Nb15 Patient Cocktail Draft
Jupyter notebook from the Metagenome-Prioritized Phage Cocktails for Crohn's Disease and IBD project.
NB15 — UC Davis Per-Patient Profile + Cocktail Draft (Pillar 5 Opener)¶
Project: ibd_phage_targeting — Pillar 5 first notebook
Depends on: NB02 patient ecotype assignments; crohns_patient_demographics; NB05 actionable Tier-A; NB06 H2d ecotype-specific modules; NB09c bile-acid coupling cost; NB10a species-vs-strain mediation; NB12-NB14 phage-availability stratification
Purpose¶
Assemble per-patient profiles for 23 UC Davis CD patients combining all Pillar 1-4 evidence and produce per-patient cocktail drafts with concrete phage components + caveats per patient.
Inputs (per patient)¶
- NB02 ecotype assignment — E0 / E1 / E3 / mixed (patient 6967 longitudinal E1↔E3)
- Demographics — age, sex, Montreal Classification, fecal calprotectin, CRP, medication
- Tier-A pathobiont presence — Kuehl_WGS Kaiju sample-level abundance for the 6 NB05 actionable Tier-A core species (presence threshold ≥0.001 relative abundance)
- NB05 Tier-A score + Pillar-3 mechanism profile (iron / BA-coupling-cost / mediation) + Pillar-4 phage availability (Tier-1 clinical / Tier-2 lytic / temperate-only / GAP / GAP+HIGH-cost)
Cocktail-design logic¶
Per patient × Tier-A target:
- Filter to ecotype-specific priority targets (NB06 H2d ecotype modules: E1_CD module 0 has H. hathewayi + F. plautii + E. bolteae + E. lenta + M. gnavus; E3_CD module 1 has E. lenta + H. hathewayi + E. coli + M. gnavus)
- Filter to targets actually present in patient (Kuehl_WGS abundance > 0.001)
- Apply Pillar-3 mechanism profile (iron + BA-coupling cost + mediation)
- Apply Pillar-4 phage availability (concrete components for E. coli (NB13 5-phage cocktail), E. bolteae (PMBT24), E. lenta (PMBT5); GAP/temperate-only flagged for alternative strategies)
- Generate caveats based on calprotectin (active vs quiescent), medication (steroid-sparing context), Pillar-4 GAP count, F. plautii BA-cost, mixed-ecotype longitudinal stability
Per plan v1.9 no raw reads.
In [1]:
# See run_nb15.py for full source.
§0. Load UC Davis per-patient profile inputs¶
In [2]:
# NB02 ecotype + demographics + Kuehl_WGS taxa + NB05 actionable + NB12 phage-availability
## §0. Load UC Davis per-patient profile inputs NB02 patient summary: 23 patients × 11 cols crohns_patient_demographics: 23 patients × 13 cols Kuehl_WGS: 1608 rows × 26 samples × 263 taxa Kuehl_WGS sample → participant: 29 NB05 actionable Tier-A: 6 species NB12 phage targetability: 71 species × 8 cols
§1. Per-patient Tier-A pathobiont abundance¶
In [3]:
# Kuehl_WGS sample-level Kaiju → per-patient Tier-A abundance + presence
## §1. Per-patient Tier-A pathobiont abundance Kuehl patient_ids extracted: 22 unique (from sample_id pattern) Actionable Tier-A names: ['Hungatella hathewayi', 'Mediterraneibacter gnavus', 'Escherichia coli', 'Eggerthella lenta', 'Flavonifractor plautii', 'Enterocloster bolteae'] Per-patient × Tier-A abundance matrix: (22, 6) Presence/absence (fraction of patients with abundance > 0): Hungatella hathewayi: 19/22 patients Mediterraneibacter gnavus: 21/22 patients Escherichia coli: 8/22 patients Eggerthella lenta: 16/22 patients Flavonifractor plautii: 18/22 patients Enterocloster bolteae: 19/22 patients
§2. Per-patient profile master table¶
In [4]:
# Merge ecotype + demographics + Tier-A presence + n_actionable_targets per patient
## §2. Build per-patient profile master table
Per-patient profile shape: (23, 27)
Per-ecotype distribution:
E0: 7; E1: 9; E3: 6; mixed: 1; nan: 0
Mean # actionable targets per patient by ecotype:
mean min max count
final_ecotype
0 3.857143 2 6 7
1 5.333333 5 6 9
3 3.500000 0 6 6
mixed 5.000000 5 5 1
§3. Per-patient cocktail draft (with rationale + caveats)¶
In [5]:
# Apply Pillar-1-4 framework to produce per-patient cocktail composition
## §3. Per-patient cocktail draft
Per-patient cocktail drafts (subset for first 10 patients):
=== Patient 1112 | E3 | calpro=41.0 | A2L3B1 | med=Infliximab,7.5mg/kg ev. 8 weeks ===
Tier-A targets present: 3 / 4
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Concrete phage components: ['PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• Low calprotectin (41) — disease quiescent, cocktail may not be needed
• 1 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 1317 | E1 | calpro=274.0 | L1 B2 | med=Budesonide 6mg daily ===
Tier-A targets present: 5 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)', 'PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
• On steroid — temporary control; phage cocktail as steroid-sparing strategy
=== Patient 1406 | E1 | calpro=nan | nan | med=nan ===
Tier-A targets present: 5 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)', 'PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• No calprotectin — disease activity not quantified
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 1460 | E1 | calpro=7.0 | L3 B1 | med=Rinvoq 45 mg. Protonix 40mg ===
Tier-A targets present: 5 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)', 'PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• Low calprotectin (7) — disease quiescent, cocktail may not be needed
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 1492 | E1 | calpro=9.0 | L3B1 | med=Skyrizi 360mg every 8 weeks ===
Tier-A targets present: 5 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)', 'PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• Low calprotectin (9) — disease quiescent, cocktail may not be needed
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 1676 | E0 | calpro=954.0 | L2 B1 | med=Infliximab injection 575 mg every 4 weeks ===
Tier-A targets present: 1 / 2
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
=== Patient 1773 | E1 | calpro=668.0 | L2 B1 | med=Budesonide 9mg daily ===
Tier-A targets present: 5 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)', 'PMBT5 (siphovirus; co-monitor BA pool)']
Caveats:
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
• On steroid — temporary control; phage cocktail as steroid-sparing strategy
=== Patient 1835 | E1 | calpro=3340.0 | L3B1 | med=Infliximab injection 300mg mg every 4 weeks ===
Tier-A targets present: 4 / 5
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Enterocloster bolteae | Tier-2 lytic | BA-cost: moderate | mixed
Flavonifractor plautii | GAP+HIGH-cost | BA-cost: HIGHEST | species-abundance
Concrete phage components: ['PMBT24 (virulent, 99,962 bp Kielviridae; co-monitor BA pool)']
Caveats:
• High calprotectin (3340) — active disease, cocktail-design priority
• 2 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 2708 | E3 | calpro=3.0 | L3 B2 B3 | med=Skyrizi 360mg/2.4 ml every 8 weeks ===
Tier-A targets present: 4 / 4
Eggerthella lenta | Tier-2 lytic | BA-cost: moderate | drug-metabolism
Hungatella hathewayi | GAP | BA-cost: low | species-abundance
Escherichia coli | Tier-1 clinical | BA-cost: low | strain-content
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
Concrete phage components: ['PMBT5 (siphovirus; co-monitor BA pool)', 'DIJ07_P2 (Phapecoctavirus, NB13 broadest host range 63.8%)', 'LF73_P1 (Tequatrovirus, Straboviridae)', 'AL505_Ev3', '55989_P2', 'LF110_P2']
Caveats:
• Low calprotectin (3) — disease quiescent, cocktail may not be needed
• 1 target(s) in Pillar-4 phage GAP — external DB query priority
=== Patient 3500 | E0 | calpro=94.0 | L2 B1 | med=SKYRIZI 360mg/2.4mL injector every 8 weeks ===
Tier-A targets present: 1 / 2
Mediterraneibacter gnavus | temperate-only | BA-cost: low | species-abundance
§4. Cocktail summary statistics¶
In [6]:
# Per-ecotype cocktail composition + per-target prescribing rate
## §4. Cocktail summary statistics
Per-ecotype cocktail summary:
ecotype n_patients mean_targets_per_patient n_with_cocktail n_with_concrete_phage_component
E0 7 1.71 7 0
E1 9 4.89 9 9
E3 6 2.33 5 4
Emixed 1 3.00 1 1
Per-target prescribing rate (across all 23 patients):
species n_present pct_present
Mediterraneibacter gnavus 21 91.3
Hungatella hathewayi 19 82.6
Enterocloster bolteae 19 82.6
Flavonifractor plautii 18 78.3
Eggerthella lenta 16 69.6
Escherichia coli 8 34.8
§5. Verdict + figure¶
In [7]:
# 3-panel: per-patient × Tier-A presence heatmap + per-ecotype prescribing rate + per-patient calprotectin × n_targets scatter
## §5. Verdict + figure
{
"date": "2026-04-25",
"plan_version": "v1.9",
"test": "NB15 \u2014 UC Davis per-patient profile + cocktail draft (Pillar 5 opener)",
"n_patients": 23,
"ecotype_distribution": {
"1": 9,
"0": 7,
"3": 6,
"mixed": 1
},
"mean_actionable_tier_a_per_patient": 4.39,
"patients_with_concrete_phage_component": 14,
"cocktail_summary_by_ecotype": [
{
"ecotype": "E0",
"n_patients": 7,
"mean_targets_per_patient": 1.71,
"n_with_cocktail": 7,
"n_with_concrete_phage_component": 0
},
{
"ecotype": "E1",
"n_patients": 9,
"mean_targets_per_patient": 4.89,
"n_with_cocktail": 9,
"n_with_concrete_phage_component": 9
},
{
"ecotype": "E3",
"n_patients": 6,
"mean_targets_per_patient": 2.33,
"n_with_cocktail": 5,
"n_with_concrete_phage_component": 4
},
{
"ecotype": "Emixed",
"n_patients": 1,
"mean_targets_per_patient": 3.0,
"n_with_cocktail": 1,
"n_with_concrete_phage_component": 1
}
],
"narrative": "23 UC Davis CD patients \u00d7 per-patient profile (ecotype + calprotectin + Montreal + medication + Tier-A pathobiont presence). E0=7, E1=9, E3=6, mixed=1 (patient 6967), nan=0. Mean 4.4 actionable Tier-A targets per patient. Per-patient cocktail draft composition stratified by ecotype-specific module (NB06 H2d) + phage-availability tier (NB12-NB14) + bile-acid coupling cost (NB09c)."
}
Wrote /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/data/nb15_per_patient_cocktail_draft.tsv (74 rows = patients × targets)
Wrote /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/figures/NB15_patient_cocktail_draft.png
§6. Interpretation¶
Headline: 14 of 23 UC Davis patients have a concrete phage cocktail draft; F. plautii BA-cost is the dominant per-patient design constraint in 9 E1 patients¶
Per-ecotype cocktail summary¶
| Ecotype | n patients | Mean targets per patient | n with cocktail | n with concrete phage |
|---|---|---|---|---|
| E0 | 7 | 1.71 | 7 | 0 (priority targets H. hathewayi + M. gnavus are GAP / temperate-only) |
| E1 | 9 | 4.89 | 9 | 9 (all E1 patients have full Tier-A module + concrete cocktails) |
| E3 | 6 | 2.33 | 5 | 4 |
| Mixed (6967) | 1 | 3.0 | 1 | 1 |
| Total | 23 | 4.39 | 22 | 14 |
Per-target prescribing rate (across all 23 patients):
- M. gnavus: 91 % (21/23)
- H. hathewayi: 83 % (19/23)
- E. bolteae: 83 % (19/23)
- F. plautii: 78 % (18/23)
- E. lenta: 70 % (16/23)
- ***E. coli*: 35 %** (8/23) — much lower carriage in UC Davis cohort
***E. coli* is present in only 8 of 23 patients** — interesting given E. coli AIEC is canonically a CD-associated species. This suggests:
- Many UC Davis patients are not AIEC-carriers despite active CD — E. coli phage cocktail (NB13 5-phage) only applicable to ~35 % of cohort.
- H. hathewayi / M. gnavus are nearly universal in this cohort — their phage GAP (NB12) is the dominant unmet need.
- Per-patient phage cocktail has 2 components for most patients: PMBT24 (E. bolteae) + PMBT5 (E. lenta). Patients carrying E. coli additionally get the NB13 5-phage AIEC cocktail.
Per-patient stratification¶
All 23 patients fall into one of 4 cocktail-design categories:
Category A — Active disease + multiple targets (high priority Pillar-5): patients 3701 (calp 8000, 5 targets, E1), 1835 (calp 3340, 4 targets, E1), 5843 (calp 2150, 5 targets, E0), 4091 (calp 1820, 5 targets, E0), 5115 (calp 1970, 5 targets, E3), 9911 (calp 1670, 5 targets, E1), 6434 (calp 251, 6 targets, E0). All E1 patients in this category have full 5-phage E. coli cocktail + PMBT24 + PMBT5 with caveats on F. plautii BA-cost.
Category B — Active disease + few targets: patients in E0 with limited pathobiont burden. May not need cocktail per ecotype-specific rationale (E0 is the diverse-commensal ecotype).
Category C — Quiescent disease: patients with calp <50 (1112, 2708, 1492, 1460) — cocktail may not be needed despite Tier-A target presence; reserve cocktail for flares.
Category D — Patient 6967 (mixed ecotype): longitudinal E1 ↔ E3 drift; central per-patient stability test for Pillar 5 dosing strategy. State-dependent cocktail (rebalance per visit ecotype) or E1+E3 union cocktail.
F. plautii BA-cost is the dominant E1 design constraint¶
***F. plautii* is present in 78 % of patients** (18/23) AND has the HIGHEST bile-acid coupling cost (NB09c §13: active 7α-dehydroxylator). All 9 E1 patients carry F. plautii. This is the dominant per-patient design constraint in the E1 ecotype:
- Targeting F. plautii via phage shifts BA pool toward inflammatory primary tauro-conjugated forms
- Alternative: deprioritize F. plautii from cocktail; co-administer UDCA / BA-binding agent if BA pool monitoring shows shift
- Phage-availability is also GAP for F. plautii (not in ref_phage_biology) — double penalty makes F. plautii a "leave alone" target for Pillar 5
Pillar-4 GAP count per patient¶
For E1 patients carrying full Tier-A module: 2 of 5 targets are Pillar-4 GAP (H. hathewayi no phages; F. plautii not in ref + HIGHEST BA-cost). Cocktail strategy:
- Direct phage targeting for 2-3 species: E. coli (if present), E. bolteae (PMBT24), E. lenta (PMBT5)
- Alternative for 2 species: H. hathewayi (GAG-degrading enzyme inhibitors per ref_phage_biology therapeutic_targets); F. plautii (deprioritize + BA-binding co-therapy)
- Limited for 1 species: M. gnavus (temperate-only; lytic-locked engineering OR biochemical glucorhamnan-synthesis target)
This is a 3-strategy hybrid cocktail for E1 patients, not a pure phage cocktail. The Pillar-4 GAP species require non-phage alternatives — INPHARED + IMG/VR external DB queries are the highest-priority Pillar-4 follow-up before clinical translation.
Pillar 5 hand-off — concrete deliverables for each patient class¶
| Patient class | n | Cocktail strategy |
|---|---|---|
| E1 + active calp + E. coli present | ~3 | Full Tier-1 cocktail: NB13 5-phage E. coli + PMBT24 + PMBT5 + alternatives for F. plautii / H. hathewayi / M. gnavus |
| E1 + active calp + no E. coli | ~5 | Tier-2 cocktail: PMBT24 + PMBT5 + alternatives |
| E1 + quiescent | ~1-2 | Reserve cocktail for flares; monitor calp |
| E0 + any calp | 7 | Limited cocktail (priority targets H. hathewayi / M. gnavus are GAP/temperate); consider deprioritizing all but very-active E0 patients |
| E3 + E. coli present | ~3 | Full Tier-1 cocktail (E. coli AIEC subset) + PMBT5 + alternatives |
| E3 + no E. coli | ~3 | Limited cocktail (E. lenta + H. hathewayi targets only) |
| Mixed (6967) | 1 | State-dependent dosing (per-visit ecotype rebalancing); central per-patient stability test |
Limitations¶
- 23-patient cohort is small for per-patient cocktail validation; results are exemplars/templates rather than statistically robust per-patient recommendations.
- Kuehl_WGS uses Kaiju classifier (not MetaPhlAn3); known classifier-mismatch asymmetry from NB02 — Tier-A presence calls have lower confidence than CMD MetaPhlAn3 calls.
- No per-patient bile-acid measurements; BA-coupling-cost annotation is ecotype-level, not per-patient.
- No per-patient pks-island / Yersiniabactin diagnostic for AIEC strain-resolution within the 8 E. coli-positive patients; cocktail recommendation for these patients assumes AIEC subset prevalence per Dogan 2014/Dubinsky 2022 but cannot be patient-specifically validated.
- F. plautii BA-cost is HIGHEST but not patient-quantified; deprioritization is precautionary based on NB09c paired-sample mechanism finding.
Outputs¶
data/nb15_patient_profile.tsv— 23 patients × full per-patient profile (ecotype, demographics, Tier-A presence, n_actionable_targets)data/nb15_per_patient_cocktail_draft.tsv— long-format per-patient × per-target cocktail breakdowndata/nb15_pillar5_cocktail_verdict.json— formal verdict + per-ecotype summaryfigures/NB15_patient_cocktail_draft.png— 3-panel: presence heatmap + ecotype prescribing rate + calp × n_targets scatter