◆ Microbial Discovery Forge AI Co-Scientist

• Co-Scientist
• Observatory
  • Projects
  • Research Areas
  • Collections
  • Data Explorer
  • Discoveries
  • Community
• Atlas
  • BERIL Atlas
  • Skills
  • Memory: Pitfalls
  • Memory: Performance
  • Research Ideas
• About

Log in with ORCiD

Projects / Metagenome-Prioritized Phage Cocktails for Crohn's Disease and IBD / NB12_phage_targetability.ipynb

Nb12 Phage Targetability

Jupyter notebook from the Metagenome-Prioritized Phage Cocktails for Crohn's Disease and IBD project.

Back to Project

NB12 — Pathobiont × phage targetability matrix (Pillar 4 opener)

Project: ibd_phage_targeting — Pillar 4 first notebook Depends on: ref_phage_biology (12-organism literature-curated phage synthesis); NB05 actionable Tier-A scoring; Pillar-3 per-target mechanism profile (iron / bile-acid / mediation)

Purpose

Build a per-pathobiont phage-targetability profile (Tier-B in the project's 4-tier rubric) for the 6 actionable Tier-A core species, combining NB05 Tier-A scoring (criteria A3-A6) with literature-curated phage availability + Pillar-3 mechanism profile to produce per-target priority classification for Pillar-5 cocktail design.

Method

Per plan v1.9 (no raw reads):

1. ref_phage_biology (12 organisms, indicator_taxa_literature_review) — literature-curated phage info per Tier-1/Tier-2 pathobiont (known_phages, therapeutic_targets, lifestyle, clinical trial status)
2. NB05 Tier-A scoring (71 candidates × A3-A6 criteria + total_score)
3. Pillar-3 per-target mechanism profile (iron specialization, bile-acid coupling cost, mechanism mediation — from REPORT §closure cocktail-design table)
4. Phage-availability score (Tier-B): 0-3 ordinal scale
   • 0 = no known phages OR only historical phage-like particles
   • 1 = temperate / prophage only, or limited characterization
   • 2 = lytic phage(s) characterized in literature, but not in clinical trials
   • 3 = clinical trial / commercial cocktail OR published efficacy data

External phage DB queries (PhageFoundry BERDL, INPHARED, IMG/VR, NCBI Phage RefSeq, PhagesDB) flagged for follow-up — BERDL Spark auth currently blocks direct PhageFoundry query, so this NB12 establishes the curated-literature-based foundation.

Tests

1. Per-pathobiont phage-availability score
2. Combined Tier-A × Tier-B × Pillar-3 mechanism profile → Pillar-5 cocktail-design priority class
3. Coverage gap analysis: which actionable Tier-A core have NO phage-therapy options?
4. External phage DB references for follow-up

# See run_nb12.py for full source.

§0. Load ref_phage_biology + NB05 Tier-A scoring + Pillar-3 mechanism profile

# 12 organisms in ref_phage_biology; 6 actionable + 9 Tier-B candidates from NB05

## §0. Load ref_phage_biology + NB05 Tier-A scoring + Pillar-3 mechanism profile
ref_phage_biology: 12 organisms × 8 columns
  Tier breakdown: {'Tier1': 6, 'Tier2': 6}

NB05 actionable Tier-A core: 6 species
  Hungatella hathewayi: total_score=4.0
  Mediterraneibacter gnavus: total_score=3.8
  Escherichia coli: total_score=3.6
  Eggerthella lenta: total_score=3.3
  Flavonifractor plautii: total_score=3.3
  Enterocloster bolteae: total_score=2.8

NB05 Tier-B candidates (score 2.2-2.4):
  Enterocloster asparagiformis: total_score=2.4
  Streptococcus salivarius: total_score=2.4
  Enterocloster citroniae: total_score=2.4
  Enterocloster clostridioformis: total_score=2.4
  Blautia coccoides: total_score=2.4
  Veillonella atypica: total_score=2.2
  Streptococcus parasanguinis: total_score=2.2
  Actinomyces oris: total_score=2.2
  Veillonella parvula: total_score=2.2

§1. Per-pathobiont phage-availability scoring (Tier-B)

# 0-3 ordinal scale based on lifestyle + clinical status

## §1. Per-pathobiont phage-availability score

Phage-availability scoring (top 20 by Tier-A score):
                       species  tier_a_score_nb05  actionable_tier_a  phage_score_tier_b               lifestyle_dominant
          Hungatella hathewayi                4.0               True                 0.0                             none
     Mediterraneibacter gnavus                3.8               True                 1.0                        temperate
              Escherichia coli                3.6               True                 3.0                   lytic_clinical
             Eggerthella lenta                3.3               True                 2.0                            lytic
        Flavonifractor plautii                3.3               True                 NaN unknown_not_in_ref_phage_biology
         Enterocloster bolteae                2.8               True                 2.0                            lytic
  Enterocloster asparagiformis                2.4              False                 1.0                   temperate_only
      Streptococcus salivarius                2.4              False                 NaN unknown_not_in_ref_phage_biology
       Enterocloster citroniae                2.4              False                 NaN unknown_not_in_ref_phage_biology
Enterocloster clostridioformis                2.4              False                 NaN unknown_not_in_ref_phage_biology
             Blautia coccoides                2.4              False                 NaN unknown_not_in_ref_phage_biology
           Veillonella atypica                2.2              False                 NaN unknown_not_in_ref_phage_biology
   Streptococcus parasanguinis                2.2              False                 NaN unknown_not_in_ref_phage_biology
              Actinomyces oris                2.2              False                 NaN unknown_not_in_ref_phage_biology
           Veillonella parvula                2.2              False                 NaN unknown_not_in_ref_phage_biology
           Hungatella symbiosa                2.1              False                 NaN unknown_not_in_ref_phage_biology
              Blautia wexlerae                2.1              False                 NaN unknown_not_in_ref_phage_biology
  Bacteroides cellulosilyticus                2.0              False                 NaN unknown_not_in_ref_phage_biology
            Veillonella dispar                2.0              False                 NaN unknown_not_in_ref_phage_biology
           Streptococcus mitis                1.9              False                 NaN unknown_not_in_ref_phage_biology

§2. Combined Tier-A × Tier-B × Pillar-3 priority for Pillar-5 cocktail design

# Per-actionable-target full profile + Pillar-5 priority class

## §2. Combined per-target priority for Pillar-5 cocktail design

Combined per-actionable-target priority (NB05 actionable Tier-A only):
  Hungatella hathewayi                | NB05=4.0 | phage=0 | none
    iron: none; BA-cost: low; mediation: species-abundance + within-carrier-metabolic-shift
    Pillar-5 class: Tier-1 phage target (highest NB05 score; low BA cost)

  Mediterraneibacter gnavus           | NB05=3.8 | phage=1 | temperate
    iron: none; BA-cost: low; mediation: species-abundance (mucin-glucorhamnan; Henke 2019)
    Pillar-5 class: Tier-1 phage target (low BA cost; mucin mechanism)

  Escherichia coli                    | NB05=3.6 | phage=3 | lytic_clinical
    iron: dominant (Yersiniabactin/Enterobactin/Colibactin); BA-cost: low; mediation: strain-content (AIEC subset specific)
    Pillar-5 class: Tier-1 phage target with strain-resolution requirement

  Eggerthella lenta                   | NB05=3.3 | phage=2 | lytic
    iron: none; BA-cost: moderate (partial 7α-dehydroxylator); mediation: drug-metabolism (Koppel 2018 Cgr2)
    Pillar-5 class: Tier-2 phage target (moderate BA cost; non-BGC drug-metabolism mechanism)

  Flavonifractor plautii              | NB05=3.3 | phage=? | unknown_not_in_ref_phage_biology
    iron: weak; BA-cost: HIGHEST (active 7α-dehydroxylator); mediation: species-abundance (NB10a F. plautii informative null)
    Pillar-5 class: Tier-2 phage target with HIGHEST BA-coupling cost — depletion shifts BA pool toward primary tauro-conjugated forms; consider co-administering UDCA / BA-binding agent

  Enterocloster bolteae               | NB05=2.8 | phage=2 | lytic
    iron: none; BA-cost: moderate (active 7α-dehydroxylator, NB09c × deoxycholate=+0.17); mediation: mixed
    Pillar-5 class: Tier-2 phage target (moderate BA cost)

§3. Coverage gap analysis

# Pathobionts with phage-availability score = 0 or unknown

## §3. Coverage gap analysis

Pathobionts with phage_score = 0 or unknown (no actionable phage options in current scope):
  Hungatella hathewayi                [ACTIONABLE] (NB05=4.0)
  Flavonifractor plautii              [ACTIONABLE] (NB05=3.3)
  Streptococcus salivarius            (NB05=2.4)
  Enterocloster citroniae             (NB05=2.4)
  Enterocloster clostridioformis      (NB05=2.4)
  Blautia coccoides                   (NB05=2.4)
  Veillonella atypica                 (NB05=2.2)
  Streptococcus parasanguinis         (NB05=2.2)
  Actinomyces oris                    (NB05=2.2)
  Veillonella parvula                 (NB05=2.2)
  Hungatella symbiosa                 (NB05=2.1)
  Blautia wexlerae                    (NB05=2.1)
  Bacteroides cellulosilyticus        (NB05=2.0)
  Veillonella dispar                  (NB05=2.0)
  Streptococcus mitis                 (NB05=1.9)
  Streptococcus oralis                (NB05=1.9)
  Lactococcus lactis                  (NB05=1.9)
  Gordonibacter pamelaeae             (NB05=1.9)
  Streptococcus vestibularis          (NB05=1.9)
  Eisenbergiella massiliensis         (NB05=1.9)
  Bifidobacterium dentium             (NB05=1.9)
  Alistipes onderdonkii               (NB05=1.9)
  Streptococcus sanguinis             (NB05=1.9)
  Streptococcus mutans                (NB05=1.9)
  Streptococcus australis             (NB05=1.9)
  Pseudoflavonifractor sp. An184      (NB05=1.9)
  Anaerotruncus colihominis           (NB05=1.7)
  Bifidobacterium breve               (NB05=1.7)
  Gemella sanguinis                   (NB05=1.7)
  Sellimonas intestinalis             (NB05=1.7)
  Streptococcus thermophilus          (NB05=1.7)
  Intestinibacter bartlettii          (NB05=1.7)
  Clostridium paraputrificum          (NB05=1.7)
  Butyricicoccus pullicaecorum        (NB05=1.7)
  Enterococcus faecalis               (NB05=1.7)
  Erysipelatoclostridium ramosum      (NB05=1.6)
  Erysipelatoclostridium innocuum     (NB05=1.6)
  Clostridium spiroforme              (NB05=1.5)
  Prevotella copri                    (NB05=1.5)
  Actinomyces odontolyticus           (NB05=1.4)
  Actinomyces sp. HPA0247             (NB05=1.4)
  Adlercreutzia caecimuris            (NB05=1.4)
  Collinsella intestinalis            (NB05=1.4)
  Blautia sp. CAG:257                 (NB05=1.4)
  Actinomyces sp. ICM47               (NB05=1.4)
  Dorea sp. CAG:317                   (NB05=1.4)
  Enterocloster aldenensis            (NB05=1.4)
  Clostridium scindens                (NB05=1.3)
  Aeriscardovia aeriphila             (NB05=1.2)
  Actinomyces sp. HMSC035G02          (NB05=1.2)
  Turicimonas muris                   (NB05=1.2)
  Roseburia sp. CAG:471               (NB05=1.2)
  Veillonella infantium               (NB05=1.2)
  Firmicutes bacterium CAG:424        (NB05=1.2)
  Clostridium sp. CAG:242             (NB05=1.2)
  Dielma fastidiosa                   (NB05=1.2)
  Anaerostipes hadrus                 (NB05=1.2)
  Butyricimonas synergistica          (NB05=1.0)
  Eubacterium sp. CAG:251             (NB05=1.0)
  Anaerostipes caccae                 (NB05=1.0)
  Bacteroides stercoris               (NB05=1.0)
  Eubacterium sp. CAG:38              (NB05=1.0)
  Eubacterium sp. CAG:274             (NB05=1.0)
  Romboutsia ilealis                  (NB05=1.0)
  Holdemanella biformis               (NB05=1.0)
  Roseburia faecis                    (NB05=0.5)

# actionable Tier-A core with phage coverage gap: 2

### Specific gaps in actionable Tier-A core:

- ***F. plautii***: NOT in ref_phage_biology curated set; literature search required (INPHARED / IMG/VR / NCBI Phage RefSeq / PhagesDB). Tier-2 priority for Pillar-5 cocktail design due to HIGH BA-coupling cost; the bile-acid-cost annotation is more important than phage availability for this species — first decision is whether to target F. plautii at all (BA pool consequence) before phage selection.

- ***H. hathewayi***: ref_phage_biology entry says "No specific phages identified; historical phage-like particles but no plaques" — CRITICAL COVERAGE GAP for the highest-NB05-scored Tier-A. External DB query required: search INPHARED / IMG/VR for any Hungatella-host phages. Also possible: target the GAG-degrading enzyme directly (per ref_phage_biology therapeutic_targets) rather than via phage.

- ***M. gnavus*** (= R. gnavus): all 6 known phages are TEMPERATE — phage therapy is structurally limited because temperate phages can confer host fitness benefits (lysogeny) rather than reliably lyse. Therapeutic strategies: (a) engineer lytic-locked variants of the temperate phages; (b) target glucorhamnan synthesis biochemically (per ref_phage_biology therapeutic_targets) rather than via phage.

§4. External phage DB references for follow-up

# PhageFoundry, INPHARED, IMG/VR, NCBI Phage RefSeq, PhagesDB

## §4. External phage DB references (out-of-BERDL — Pillar 4 follow-up)

- **PhageFoundry (BERDL)**
  Access: BERDL Spark Connect — phagefoundry_strain_modelling, phagefoundry_ecoliphages_genomedepot, phagefoundry_klebsiella_*, phagefoundry_acinetobacter_*, phagefoundry_paeruginosa_*, phagefoundry_pviridiflava_*
  Coverage: E. coli direct (phagefoundry_ecoliphages_genomedepot); K. oxytoca direct (phagefoundry_klebsiella_*). Other Tier-A core (gut commensal hosts) not directly covered by current PhageFoundry collections.
  Status: BLOCKED at NB12 execution: BERDL auth token in .env stale (KBASE_AUTH_TOKEN reports invalid). Refresh token + re-query as Pillar-4 follow-up.

- **Millard lab INPHARED**
  Access: http://millardlab.org/phages/inphared/ (downloadable phage genome annotations + host predictions)
  Coverage: Comprehensive — ~25K phage genomes with GenBank-quality annotations. Host predictions via BLAST + phylogenetic placement. Best single source for phage availability across all Tier-A pathobionts.
  Status: Out-of-BERDL — manual download + parse required. Promote to NB12-followup for the 3 actionable Tier-A coverage gaps (F. plautii, H. hathewayi, R. gnavus lytic alternatives).

- **IMG/VR v4**
  Access: https://genome.jgi.doe.gov/portal/IMG_VR/IMG_VR.home.html (~3M uncultivated viral genomes from metagenomes, with host predictions via CRISPR spacer matches and BLAST)
  Coverage: Strongest for uncultivated phages of gut-anaerobe hosts where culturing-based isolation has failed (H. hathewayi, F. plautii). UViGs in IMG/VR may include phages with CRISPR-spacer-derived host predictions.
  Status: Out-of-BERDL — JGI Portal API or direct download. Promote to NB12-followup.

- **NCBI Phage Virus RefSeq**
  Access: NCBI Virus / RefSeq Viral; ~5K curated phage reference genomes
  Coverage: Curated subset — covers well-studied phages (E. coli, Klebsiella, Salmonella, Pseudomonas) but limited gut-anaerobe coverage.
  Status: Out-of-BERDL — straightforward NCBI E-utils query.

- **PhagesDB**
  Access: https://phagesdb.org/ (Mycobacterium-phage focused; ~25K isolated phages)
  Coverage: Mostly mycobacteriophages; not directly relevant for IBD pathobionts.
  Status: Low priority for this project.

§5. Verdict + figure

# 2-panel: Tier-A × Tier-B scatter + per-actionable Pillar-5 priority bar

## §5. Save outputs + verdict
{
  "date": "2026-04-25",
  "plan_version": "v1.9",
  "test": "NB12 \u2014 Pathobiont \u00d7 phage targetability matrix (Pillar 4 opener)",
  "n_actionable_tier_a": 6,
  "n_phage_score_3_clinical": 1,
  "n_phage_score_2_lytic_literature": 2,
  "n_phage_score_1_temperate_or_limited": 1,
  "n_phage_score_0_gap": 1,
  "phage_clinical_tier_a": [
    "Escherichia coli"
  ],
  "phage_lytic_literature_tier_a": [
    "Eggerthella lenta",
    "Enterocloster bolteae"
  ],
  "phage_limited_tier_a": [
    "Mediterraneibacter gnavus"
  ],
  "phage_gap_tier_a": [
    "Flavonifractor plautii",
    "Hungatella hathewayi"
  ],
  "pillar4_pillar5_handoff_note": "Phage availability stratifies the 6 actionable Tier-A core into 4 classes: clinical-trial-stage (E. coli AIEC); lytic-literature (E. lenta, E. bolteae); temperate-limited (M. gnavus); coverage-gap (H. hathewayi, F. plautii). The 2 coverage-gap targets are the highest-NB05-scored species and require external phage DB queries (INPHARED + IMG/VR) as Pillar-4 follow-up. F. plautii additionally carries the highest BA-coupling cost \u2014 phage targeting may be deprioritized in favor of bile-acid-pool monitoring or biochemical-target alternatives. M. gnavus temperate-only constraint may require either lytic-locked phage engineering or biochemical glucorhamnan-synthesis targeting.",
  "limitations": [
    "BERDL Spark auth blocked at NB12 execution (KBASE_AUTH_TOKEN stale); PhageFoundry collections not directly queried. Refresh token + cross-check ref_phage_biology curated synthesis as a Pillar-4 follow-up.",
    "External phage DB queries (INPHARED / IMG/VR / NCBI Phage RefSeq) are out-of-BERDL and not run in NB12. Coverage gaps for F. plautii / H. hathewayi specifically require these queries.",
    "ref_phage_biology has 12 organisms \u2014 F. plautii is NOT in the curated set (only 5 of 6 actionable Tier-A core covered).",
    "Phage-availability scoring is qualitative (0-3 ordinal scale based on lifestyle + clinical status) \u2014 quantitative coverage metrics (n_phages, host-range CDS, receptor-binding-domain diversity) require PhageFoundry / INPHARED genomic data."
  ]
}

Wrote /home/aparkin/BERIL-research-observatory-ibd/projects/ibd_phage_targeting/figures/NB12_phage_targetability.png

§6. Interpretation

Headline: Phage availability stratifies the 6 actionable Tier-A core into 4 priority classes; H. hathewayi (highest NB05) and F. plautii (highest BA-cost) are coverage gaps requiring external DB queries

Per-actionable Tier-A phage-targetability profile

Pathobiont	NB05 score	Phage score	Lifestyle	BA cost	Pillar-5 class
H. hathewayi	4.0	0	none	low	GAP: highest NB05 but no known phages — external DB query (INPHARED + IMG/VR) priority
M. gnavus	3.8	1	temperate	low	Limited: 6 known phages all temperate — lytic-locked engineering OR biochemical glucorhamnan-synthesis target as alternatives
E. coli (AIEC)	3.6	3	lytic + clinical	low	Tier-1 clinical: EcoActive cocktail (7 lytic phages, clinical trials); HER259 (FimH-targeting, attenuates virulence). Most advanced. Strain-resolution requirement (AIEC subset) per NB07b/NB08a
E. lenta	3.3	2	lytic literature	moderate	Tier-2: PMBT5 siphovirus characterized; non-BGC drug-metabolism mechanism (Koppel 2018 Cgr2) — moderate-priority target
F. plautii	3.3	0	unknown (not in ref)	HIGHEST	GAP + HIGH cost: not in ref_phage_biology; HIGHEST BA-coupling cost (active 7α-dehydroxylator). Phage targeting deprioritized in favor of bile-acid-pool monitoring or biochemical alternatives
E. bolteae	2.8	2	lytic literature	moderate	Tier-2: PMBT24 (virulent, 99,962 bp Kielviridae) — best-characterized lytic phage among gut-anaerobe Tier-A

Stratification — four phage-availability classes among 6 actionable Tier-A:

• Class 3 (clinical trial): 1 species — E. coli (EcoActive cocktail; Galtier 2017 mouse model precedent)
• Class 2 (lytic literature): 2 species — E. lenta (PMBT5), E. bolteae (PMBT24)
• Class 1 (temperate / limited): 1 species — M. gnavus (6 temperate siphoviruses; lifestyle limits therapy)
• Class 0 (gap): 2 species — H. hathewayi (no specific phages identified), F. plautii (not in ref_phage_biology)

Critical observations

1. The 2 highest-NB05-scored species (H. hathewayi 4.0, M. gnavus 3.8) have the WEAKEST phage availability. Phage-therapy success requires resolving these gaps (INPHARED / IMG/VR for H. hathewayi; lytic-locked phage engineering or biochemical alternatives for M. gnavus).

2. ***F. plautii* has both phage GAP AND highest BA-coupling cost** (NB09c §13: active 7α-dehydroxylator, depletion shifts BA pool toward inflammatory primary tauro-conjugated forms). This makes F. plautii a lowest-priority Pillar-5 target despite NB05 score 3.3 — first decision is whether to target at all (BA pool consequence) before phage selection. Alternative strategies: (a) co-administer UDCA / BA-binding agent; (b) target downstream of F. plautii (e.g., bile-acid-pool replenishment); (c) accept partial F. plautii depletion with clinical BA monitoring.

3. ***E. coli* AIEC is the highest-Pillar-5-feasibility target**: clinical-trial-stage phage cocktail (EcoActive — 7 lytic phages), low BA-coupling cost, mechanism well-characterized (NB05 §5g + NB07c §2 + NB08a §2 iron-acquisition narrative). The Pillar-4-feasibility decision for E. coli is sharpened by the AIEC strain-content requirement: target Yersiniabactin/Enterobactin/Colibactin-positive strains specifically (per NB08a) — generic E. coli phages may not deplete the right subset.

4. ***E. bolteae* + E. lenta are mid-tier targets** with lytic literature phages and moderate BA-coupling cost. Both are realistic Pillar-5 phage-cocktail components subject to BA pool monitoring.

Pillar 4 → Pillar 5 hand-off framework

The 6 actionable Tier-A core stratify into 3 Pillar-5 design strategies:

1. Direct phage targeting (Tier-1): E. coli (AIEC subset, clinical-trial cocktail) → use EcoActive or build similar 7-phage cocktail; require strain-resolution diagnostic.
2. Phage targeting with monitoring (Tier-2): E. lenta (PMBT5), E. bolteae (PMBT24) → include in cocktail with BA-pool monitoring; M. gnavus is here too if lytic-locked engineering succeeds.
3. Phage GAP — alternative strategies needed:
   • H. hathewayi: highest priority for external DB query (INPHARED / IMG/VR) — if no phages found, fall back to GAG-degrading enzyme inhibitors per ref_phage_biology therapeutic_targets.
   • F. plautii: lowest Pillar-5 priority due to highest BA-cost — consider deprioritizing or replacing phage approach with BA-binding co-therapy.
   • M. gnavus if lytic-locked engineering fails: biochemical glucorhamnan-synthesis targets (Henke 2019).

Limitations

• BERDL Spark auth blocked at NB12 execution (KBASE_AUTH_TOKEN stale) — PhageFoundry collections (phagefoundry_strain_modelling, phagefoundry_ecoliphages_genomedepot, phagefoundry_klebsiella_*) not directly queried. Refresh token + re-query as Pillar-4 follow-up. Direct PhageFoundry coverage would primarily augment E. coli (genomes + host-range CDS + receptor-binding-domain diversity) and K. oxytoca (Tier-2).
• External phage DB queries (INPHARED ~25K phages with host predictions; IMG/VR ~3M UViGs from metagenomes; NCBI Phage RefSeq ~5K curated) are out-of-BERDL and not run in NB12. The 2 actionable Tier-A coverage gaps (F. plautii, H. hathewayi) require these queries — this is the highest-priority Pillar-4 follow-up.
• ref_phage_biology has 12 organisms — F. plautii not in the curated set (5 of 6 actionable Tier-A core covered).
• Phage-availability scoring is qualitative ordinal (0-3) based on lifestyle + clinical status. Quantitative coverage metrics (n_phages, host-range CDS, receptor-binding-domain diversity, plaque burst size, pH stability) require PhageFoundry / INPHARED genomic data.

Outputs

• data/nb12_phage_targetability_matrix.tsv — per-pathobiont scoring matrix (NB05 Tier-A score + phage Tier-B score + lifestyle + Pillar-5 class)
• data/nb12_phage_targetability_verdict.json — formal verdict + Pillar-4/5 hand-off note + limitations
• figures/NB12_phage_targetability.png — 2-panel: Tier-A × Tier-B scatter + per-actionable Pillar-5 priority bar

Microbial Discovery Forge

AI co-scientist and research observatory for BERDL-scale microbial discovery.

Maintained by Paramvir S. Dehal

Observatory

• Projects
• Collections
• Discoveries

Atlas

• BERIL Atlas
• Skills
• Pitfalls
• Performance
• Research Ideas

Resources

• BERDL JupyterHub
• Co-Scientist
• How to Cite

Built on BERDL (KBase). Last updated: 2026-06-15 21:31 UTC

×