# Steal This Grant 🧬📑 ## An Open, Forkable NIH-Style Proposal — *A Precision Bacteriophage Cocktail to Prevent Necrotizing Enterocolitis in Very-Low-Birth-Weight Infants* > **License:** CC0 / public domain. Fork it, gut it, rename it, submit it. Attribution appreciated, never required. > **What this is:** (A) a complete, specific, genuinely fundable mock proposal you can adapt, and (B) a reusable, funder-agnostic template + AI prompt library so you can write your *own* phage grant for a different indication. > **What this is not:** a substitute for your IRB, your IND pre-submission meeting, your biostatistician, or your program officer. Illustrative figures are explicitly marked **[ILLUSTRATIVE]**. Budget numbers are sketches, not quotes. **How to use this document** - Reading top to bottom gives you a model R01-scale proposal. - **Part A** is the filled grant. - **Part B** is the empty skeleton + the prompt library (methodology adapted from [eseckel/ai-for-grant-writing](https://github.com/eseckel/ai-for-grant-writing)). - Search-and-replace tokens like `[INDICATION]`, `[PATHOGEN]`, `[POPULATION]` are pre-wired in Part B. --- --- # PART A — THE FILLED GRANT --- ## Title **Pre-emptive precision phage therapy against the pre-symptomatic *Klebsiella* bloom to prevent necrotizing enterocolitis in very-low-birth-weight infants (the PHAGE-NEC program)** *Alternate titles considered (kept here so you can see the title-iteration move):* - *Intercepting the bloom: a strain-resolved bacteriophage cocktail for NEC prevention* - *From metagenome to medicine: replication-rate–guided phage prophylaxis in the preterm gut* - *PREEMPT-NEC: Precision Removal of *Enterobacteriaceae* via Engineered Microbial Phage Therapy* --- ## Project Summary / Abstract Necrotizing enterocolitis (NEC) is the most lethal acquired gastrointestinal disease of prematurity, striking 5–10% of very-low-birth-weight (VLBW, <1500 g) infants, killing 20–30% of those affected, and leaving survivors with short-bowel syndrome and neurodevelopmental injury. Despite four decades of trials, no targeted therapy exists; care remains supportive, and broad-spectrum antibiotics — the current reflex — paradoxically deplete protective commensals and select for the very pathobionts implicated in disease. A precise, mechanistic target has now emerged. Strain-resolved metagenomics shows that a **gut bloom of *Klebsiella* (and related *Enterobacteriaceae*), accompanied by elevated bacterial replication rates, precedes the clinical onset of NEC by days** (Olm et al., *Sci Adv* 2019). In gnotobiotic and preterm-piglet models, **transfer of the fecal viral (bacteriophage) fraction prevents NEC**, and **UV-inactivation of that virome abolishes protection** — establishing that *intact, replication-competent phages*, not residual metabolites, carry the protective effect (Brunse et al., *ISME J* 2021; Spiegelhauer et al., *Gut Microbes* 2025). In parallel, a rationally composed **anti-*Klebsiella* phage cocktail has been shown to suppress disease-driving *Klebsiella* strains in the mammalian gut and attenuate inflammation** (Federici et al., *Cell* 2022). We propose to convert these convergent findings into a deployable preventive: a **defined, GMP-grade, host-range-matched bacteriophage cocktail** administered enterally to VLBW infants **during the pre-symptomatic *Klebsiella* bloom window**, identified by rapid strain- and replication-rate–aware surveillance. **Aim 1** builds the phage–host matching engine and a clinical-isolate biobank, and assembles a cocktail with quantified host range, resistance-suppression (via cocktail design and phage steering), and safety/manufacturability criteria. **Aim 2** establishes efficacy and mechanism in the preterm-piglet NEC model, with the UV-inactivated cocktail as the decisive mechanistic control, and defines pharmacokinetics/pharmacodynamics (phage replication on target in vivo, off-target sparing of commensals). **Aim 3** runs a phase 1b randomized, placebo-controlled safety/biomarker trial in VLBW infants, with bloom-triggered dosing, pre-specified safety stopping rules, and longitudinal strain-resolved metagenomics as the primary mechanistic readout. **Impact:** Success delivers the first *precision*, *non-antibiotic*, *commensal-sparing* preventive for NEC, a validated bloom-surveillance pipeline reusable across neonatal pathobionts, and an open phage-matching framework. Because every reagent, protocol, and analysis pipeline is released openly, the program is designed to catalyze a field, not just a product. --- ## Specific Aims *(one page)* **The problem.** NEC remains a top cause of death in VLBW infants. Management is supportive and reactive; broad-spectrum antibiotics worsen the dysbiosis they are meant to treat. We lack any therapy that selectively removes the implicated pathobiont while sparing the protective community. **The opportunity.** Four independent lines of evidence now point to a single, actionable intervention: 1. A **pre-symptomatic *Klebsiella*/*Enterobacteriaceae* bloom with elevated in-situ replication rates precedes NEC** and is detectable by strain-resolved metagenomics (Olm 2019). 2. **Fecal-virome transfer prevents NEC** in the preterm-piglet model (Brunse 2021). 3. **UV-inactivating that virome abolishes protection**, proving the active principle is replication-competent phage, not filtrate chemistry (Spiegelhauer 2025). 4. A **defined anti-*Klebsiella* phage cocktail suppresses the target strain in vivo and dampens inflammation** (Federici 2022). **Central hypothesis.** Enteral delivery of a defined, host-range–matched bacteriophage cocktail, *timed to the pre-symptomatic Klebsiella bloom*, will selectively collapse the bloom, preserve commensal diversity, and reduce NEC incidence and severity. > **Aim 1 — Build the phage–host matching engine and assemble a defined, manufacturable anti-bloom cocktail.** > Establish a biobank of NEC-associated *Klebsiella*/*Enterobacteriaceae* clinical isolates from VLBW cohorts; characterize each by whole-genome sequencing, capsule (K-) type, and resistance/virulence content. Screen a curated phage library for host range, build a quantitative phage–host interaction matrix, and assemble a ≤6-phage cocktail meeting pre-specified criteria: ≥90% coverage of circulating bloom strains, suppression of resistant-mutant outgrowth in vitro, strictly lytic genomes free of toxin/AMR/integrase genes, and stable titer. **Outcome:** a locked, characterized cocktail + an open matching pipeline. > **Aim 2 — Establish efficacy, mechanism, and PK/PD in the preterm-piglet NEC model.** > Test the cocktail vs. vehicle and vs. **UV-inactivated cocktail** (the mechanistic control that pins causality to phage replication). Primary endpoint: NEC incidence/severity. Mechanistic endpoints: target-strain reduction, commensal sparing, mucosal inflammation, and **in vivo phage amplification on target** (replication-rate readouts). **Outcome:** causal, dose-anchored efficacy with a go/no-go threshold for clinical translation. > **Aim 3 — Phase 1b randomized, placebo-controlled safety and biomarker trial in VLBW infants with bloom-triggered dosing.** > Deploy rapid bloom surveillance; randomize bloom-positive infants to cocktail vs. placebo. Primary: safety/tolerability. Secondary/mechanistic: bloom collapse, commensal preservation, inflammatory biomarkers, and exploratory NEC incidence. **Outcome:** a safety/PK package and effect-size estimate to power a pivotal efficacy trial. **Payoff.** The first precision, commensal-sparing NEC preventive; a reusable bloom-surveillance + phage-matching platform; and a fully open toolkit to accelerate phage therapy across neonatal and beyond. --- ## Significance **Burden and unmet need.** NEC affects roughly 1 in 10 VLBW infants and is among the leading causes of death in the neonatal intensive care unit (NICU) beyond the first week of life. Mortality is 20–30% overall and exceeds 40% for surgical NEC. Survivors face intestinal failure, prolonged parenteral nutrition, cholestasis, repeated surgeries, and elevated risk of cerebral palsy and cognitive impairment. The lifetime cost per surgical-NEC survivor runs into the hundreds of thousands of dollars; the aggregate U.S. burden is in the billions annually. **Why current approaches fail.** - *Supportive care is reactive.* By the time pneumatosis intestinalis is visible on radiograph, the cascade — barrier failure, translocation, ischemic necrosis — is often irreversible. - *Antibiotics are a blunt instrument that backfires.* Prolonged empiric broad-spectrum antibiotic exposure is itself an independent risk factor for NEC: it collapses commensal diversity, removes colonization resistance, and enriches *Enterobacteriaceae* — precisely the organisms implicated in the bloom. - *Probiotics are non-specific and carry their own risks.* General probiotic supplementation has shown inconsistent benefit, lot-to-lot variability, and rare but real bacteremia/fungemia in this fragile population. They add organisms; they do not *remove the offender*. **The mechanistic pivot.** The field now has a *pre-symptomatic, strain-resolved target.* Olm et al. (2019) showed, using genome-resolved metagenomics and replication-rate inference (iRep-type measures), that NEC is preceded by a bloom of specific *Klebsiella*/*Enterobacteriaceae* strains replicating faster in situ — a signal that appears *before* clinical disease. This reframes NEC from "inflammation to be suppressed" to "a specific bacterial expansion to be *intercepted*." **Why phage, why now.** Bacteriophages are the only known antibacterial modality that is simultaneously (i) **strain-specific** — capable of removing the bloom strain while sparing *Bifidobacterium*, *Lactobacillus*, and other protective commensals; (ii) **self-amplifying** on target, so dose tracks pathogen load; and (iii) **non-antibiotic**, avoiding cross-resistance with the antibiotics this population already over-receives. The Brunse (2021) → Spiegelhauer (2025) arc is the linchpin: virome transfer prevents NEC, and **UV-killing the virome abolishes the protection** — a clean loss-of-function experiment that excludes "it's just the filtrate" and implicates *replication-competent phages* as the active agent. Federici (2022) supplies the translational template: a rationally designed anti-*Klebsiella* cocktail that suppresses a disease-driving strain in the mammalian gut and reduces inflammation, with a path through manufacturing and resistance management. **What changes if we succeed.** A positive program shifts NEC prevention from broad microbial suppression toward **targeted ecological correction**, gives neonatologists a tool that *replaces* an antibiotic reflex rather than adding to it, and validates a generalizable "surveil the bloom → match the phage → intercept" paradigm extensible to late-onset sepsis and other pathobiont-driven neonatal diseases. --- ## Innovation This program is innovative in concept, in timing, and in tooling. - **Conceptual: prevention by ecological interception, not suppression.** We do not treat established NEC; we intercept its *upstream cause* — a defined bacterial bloom — during a pre-symptomatic window. This inverts the standard reactive paradigm. - **Timing as the active ingredient.** Most antibacterial strategies are dosed by clinical event or schedule. We dose by **biology**: a strain- and replication-rate–aware surveillance trigger derived directly from the Olm 2019 signal. The *when* is as engineered as the *what*. - **A causal mechanistic control built into the design.** Few microbiome interventions can point to a loss-of-function experiment that isolates the active agent. We can: the **UV-inactivated cocktail arm** (after Spiegelhauer 2025) makes Aim 2 a true mechanism test, not just an efficacy screen. - **Precision phage matching with resistance pre-emption.** Rather than a fixed cocktail, we build a **quantitative phage–host interaction matrix** and select combinations explicitly for (i) coverage of *circulating* bloom strains and (ii) suppression of resistant-mutant escape — including evolutionary "phage steering," where resistance to one phage forces a fitness/virulence trade-off exploited by another, consistent with the Federici design logic. - **Commensal-sparing by construction.** Strictly lytic, narrow-host-range phages are selected to leave protective taxa untouched — the opposite of the antibiotic externality. - **Open by design.** Isolate genomes, the phage–host matrix, matching code, animal protocols, and trial analysis pipelines are released openly. The deliverable is not only a candidate therapeutic but a *reusable platform* the whole field can fork — which is also why this proposal itself is published as a steal-this-grant template. --- ## Approach ### Overview & experimental logic The program advances along a de-risking staircase: **(Aim 1)** make and characterize the right cocktail against the right strains; **(Aim 2)** prove it works *and why* in the most NEC-faithful animal model, with the UV control nailing causality and PK/PD anchoring dose; **(Aim 3)** show it is safe and biologically active in the target infants with bloom-triggered dosing, generating the effect size to power a pivotal trial. Surveillance (strain- and replication-rate–resolved metagenomics) is the connective tissue across all three aims. --- ### Aim 1 — Build the phage–host matching engine and assemble a defined, manufacturable anti-bloom cocktail **Rationale.** A cocktail is only as good as its match to the strains actually blooming in contemporary NICUs, and only as durable as its resistance-suppression. We therefore couple a clinical-isolate biobank to a curated phage library through a quantitative interaction matrix. **Design.** 1. **Isolate biobank.** Recover ≥300 *Klebsiella*/*Enterobacteriaceae* isolates from banked and prospectively collected VLBW stool across ≥3 NICUs, prioritizing isolates from bloom timepoints. Whole-genome sequence each; assign species, MLST, **capsule (K-) type**, and screen for AMR/virulence/toxin loci. Capture phylogenetic and capsular diversity representative of circulating bloom strains. 2. **Phage library.** Assemble ≥150 candidate lytic phages from environmental sampling (wastewater, NICU-adjacent sources), public collections, and collaborator stocks. Sequence all; **exclude any phage carrying integrase, known toxin, or AMR genes**; retain strictly lytic genomes. 3. **Interaction matrix.** Quantify host range by high-throughput efficiency-of-plating (EOP) and liquid-killing kinetics for every phage × isolate pair. Build a coverage model that maps cocktails to fraction-of-strains-covered. 4. **Resistance pre-emption.** For lead phages, isolate resistant mutants in vitro, sequence resistance loci, and measure fitness/virulence trade-offs. Select complementary phages whose combined use either covers escape mutants or **steers** resistance toward attenuated phenotypes (capsule loss, reduced colonization). 5. **Cocktail assembly & lock.** Select ≤6 phages meeting pre-specified release criteria, then **lock composition** before Aim 2. **Pre-specified cocktail release criteria (go/no-go):** | Criterion | Threshold | |---|---| | Coverage of circulating bloom strains | ≥90% by EOP ≥ 0.1 | | In vitro resistant-mutant suppression (combined cocktail) | No outgrowth over 48 h in time-kill at target MOI | | Genome safety | Strictly lytic; no integrase / AMR / toxin genes | | Manufacturability | Reaches ≥10¹⁰ PFU/mL; endotoxin removable to clinical spec | | Stability | ≤0.5 log titer loss over 6 mo at intended storage | **Expected outcomes.** A locked, fully characterized cocktail; an open phage–host interaction matrix and matching pipeline; a versioned isolate/phage biobank. **Potential pitfalls & alternatives.** - *Insufficient coverage from natural phages.* → Expand environmental sampling; add host-range-engineered or receptor-binding-protein–swapped phages; permit a 6-phage rather than 3-phage formulation. - *Rapid resistance.* → Lean on steering pairs and capsule-targeting phages whose escape mutants lose the protective capsule; pre-register an adaptive cocktail-refresh SOP. - *Strain drift between sites/years.* → Build the matching engine to be *re-run*, not one-shot; define a re-qualification cadence. --- ### Aim 2 — Establish efficacy, mechanism, and PK/PD in the preterm-piglet NEC model **Rationale.** The preterm-piglet model is the most translationally faithful NEC system (preterm delivery, enteral-feeding–induced NEC, human-relevant pathophysiology) and is the model in which virome transfer was shown to prevent NEC (Brunse 2021) and UV-inactivation to abolish it (Spiegelhauer 2025). It is therefore the right arena to test our *defined* cocktail and to run the decisive mechanistic control. **Design (preterm-piglet NEC model).** - **Arms:** (1) vehicle/placebo; (2) **active cocktail**; (3) **UV-inactivated cocktail** (matched particle dose, replication-incompetent); optional (4) **antibiotic comparator** for context. Randomized, blinded scoring. - **Dosing:** enteral, anchored to PK/PD from a dose-ranging sub-study; timed to colonization/bloom onset. - **Primary endpoint:** NEC incidence and severity (blinded macroscopic + histologic scoring). - **Mechanistic endpoints:** - **Target-strain reduction** (strain-resolved qPCR/metagenomics). - **Commensal sparing** (community diversity vs. antibiotic comparator). - **In vivo phage amplification on target** — phage titer rising with pathogen load — and **bacterial replication-rate** readouts (iRep-type) to show the bloom's growth signature is blunted. - **Mucosal inflammation** (histology, cytokines, barrier markers). - **Power:** size to detect a pre-specified absolute NEC reduction at 80% power, two-sided α 0.05; biostatistician-set group sizes; ARRIVE-compliant reporting. **The logic of the UV arm.** If the **active** cocktail prevents NEC but the **UV-inactivated** cocktail does not — despite identical particle dose — protection requires *replication-competent phage*. This is the in-program replication of the Spiegelhauer 2025 loss-of-function result and the cleanest possible causal claim for a microbiome-targeted agent. **Expected outcomes.** Causal, dose-anchored efficacy; demonstration that benefit tracks phage replication and target collapse, not filtrate chemistry; commensal sparing relative to antibiotics; a defined effective enteral dose and a clinical go/no-go. **Potential pitfalls & alternatives.** - *Model variability / low baseline NEC rate.* → Standardize induction; pre-register scoring; adequate n; consider a sensitized sub-cohort. - *Phage fails to amplify in vivo (gut transit, pH, bile).* → Encapsulation/buffering; acid protection; redosing schedule; confirm gut viability ex vivo. - *Active and UV arms both protect (filtrate effect).* → Would overturn the hypothesis; pre-specified to trigger mechanistic re-evaluation rather than silent reinterpretation — an honest off-ramp. - *Immune/endotoxin effects of particle load.* → Endotoxin-purified prep; the UV arm also controls for non-specific particle immunomodulation. --- ### Aim 3 — Phase 1b randomized, placebo-controlled safety & biomarker trial in VLBW infants with bloom-triggered dosing **Rationale.** With a locked, animal-validated cocktail and PK/PD in hand, the first-in-infant study must (i) establish safety/tolerability in VLBW infants and (ii) confirm the intended *biology* — bloom collapse with commensal preservation — while generating an effect-size estimate for a pivotal trial. Dosing is triggered by the same surveillance signal that defines the therapeutic window. **Design.** - **Population:** VLBW (<1500 g) infants; pre-specified inclusion/exclusion; informed parental consent. - **Surveillance trigger:** rapid strain- and replication-rate–aware screening of serial stool/rectal samples to identify the **pre-symptomatic *Klebsiella* bloom**; only **bloom-positive** infants are randomized (enrichment by mechanism). - **Randomization:** bloom-positive infants 1:1 to **cocktail vs. placebo**, stratified by site and gestational age; double-blind. - **Primary endpoint:** safety/tolerability (adverse events, feeding tolerance, vitals, labs; pre-specified stopping rules). - **Secondary/mechanistic endpoints:** bloom collapse (target-strain load), **commensal preservation** (diversity), inflammatory biomarkers (e.g., fecal calprotectin, circulating cytokines), phage pharmacokinetics (enteral persistence, systemic translocation screen). - **Exploratory:** NEC incidence/severity (hypothesis-generating; trial is safety-powered). - **Oversight:** independent DSMB; pre-registered; IND under FDA; pre-specified interim safety looks. **Expected outcomes.** A clean safety/tolerability and PK package in the target population; demonstration of on-target bloom collapse with commensal sparing; an effect-size estimate to power a definitive efficacy trial; a validated, deployable bloom-surveillance workflow. **Potential pitfalls & alternatives.** - *Bloom-trigger too slow for the window.* → Invest in turnaround-time reduction (targeted qPCR panel as a fast proxy for the metagenomic signal); pre-bank dosing so therapy can start within hours of a positive trigger. - *Low bloom-positive yield / slow accrual.* → Multi-site design; broaden surveillance frequency; adaptive enrollment. - *Safety signal (e.g., translocation, immune activation).* → Conservative dose-escalation; DSMB stopping rules; systemic phage and endotoxin monitoring built in. - *Regulatory novelty of a live phage biologic in neonates.* → Early and iterative FDA engagement (pre-IND), leveraging existing phage IND precedents; CMC package front-loaded in Aim 1. --- ## Timeline *(illustrative, 5-year R01-scale)* **[ILLUSTRATIVE]** ``` Year: 1 2 3 4 5 AIM 1 Biobank ███████ Phage lib ███████ Matrix/lock █████████ AIM 2 PK/PD ██████ Efficacy+UV ████████████ AIM 3 IND/CMC ████████ Surveillance ██████ Phase 1b trial ██████████████████ Cross Open data/pipeline releases ░░░░░░░░░░░░░░░░░░░░░░░ (continuous) ``` - **Y1:** Biobank + phage library + interaction matrix; cocktail lock by end of Y1/early Y2. - **Y2:** PK/PD dose-ranging; begin efficacy + UV-control study; start CMC/IND-enabling work. - **Y3:** Complete Aim 2; IND submission; stand up clinical surveillance. - **Y4–Y5:** Phase 1b enrollment, analysis, effect-size estimate; continuous open releases. --- ## Budget Justification Sketch *(illustrative, not a quote)* **[ILLUSTRATIVE]** > Replace with your institution's actual rates and your sponsored-programs office's numbers. Shown as proportional emphasis, not dollars. - **Personnel (largest share):** PD/PIs (microbial ecology/phage biology + neonatology); phage microbiologist; bioinformatician (strain-resolved metagenomics, iRep); large-animal study coordinator; clinical research coordinator(s); regulatory/CMC specialist; biostatistician. *Justification:* the program is method- and people-intensive across wet lab, animal, computation, and clinic. - **GMP/CMC manufacturing & QC:** phage amplification, purification, endotoxin removal, stability, fill-finish for Aim 2 (research grade → GMP) and Aim 3 (clinical grade). *Justification:* a defined live biologic for neonates demands rigorous CMC; front-loaded to de-risk the IND. - **Sequencing & computation:** WGS of ~300 isolates + ~150 phages; longitudinal metagenomics across animal and clinical samples; compute for matrix and replication-rate analyses. - **Preterm-piglet studies:** per-diem, surgery/feeding, histology, blinded scoring; powered group sizes across ≥3 arms + dose-ranging. - **Clinical trial costs:** IND fees, DSMB, monitoring, pharmacy, surveillance assays, biomarker panels, parental-consent infrastructure across sites. - **Open-science / dissemination:** data deposition, pipeline packaging, open protocols. - **Indirects:** per negotiated F&A rate. --- ## Vertebrate Animals (Aim 2) - **Species/justification.** Preterm piglets — the most translationally faithful NEC model (preterm delivery, enteral-feeding–induced disease, human-like intestinal pathophysiology) and the model underpinning Brunse 2021 / Spiegelhauer 2025; no non-animal system reproduces the systemic NEC phenotype. - **Numbers.** Biostatistician-determined to detect the pre-specified NEC reduction at 80% power; minimized via robust scoring and within-litter design where feasible. - **Procedures / welfare.** Standardized rearing/feeding, humane endpoints, blinded scoring, analgesia per veterinary guidance; ARRIVE-compliant reporting; IACUC-approved. - **3Rs.** *Replace* — in vitro killing/coverage assays do maximal upstream filtering so only locked candidates enter animals. *Reduce* — pre-specified power, shared controls across sub-studies. *Refine* — validated humane endpoints, early removal criteria. ## Human Subjects (Aim 3) - **Population & protections.** VLBW infants — a vulnerable population (Subpart D); minimized risk via animal-validated dose, conservative escalation, independent DSMB, pre-specified stopping rules, and IND oversight. Informed parental consent; assent not applicable. - **Scientific/clinical rationale for risk.** NEC's high mortality and the absence of any targeted preventive justify a carefully bounded first-in-infant safety study; mechanism-based enrollment (bloom-positive only) concentrates potential benefit and limits exposure. - **Design rigor.** Randomized, double-blind, placebo-controlled; pre-registered; sex/gestational-age as analysis variables; data-sharing plan honoring infant privacy. - **Inclusion across the lifespan / sex.** Both sexes enrolled and analyzed; the lifespan focus (neonates) is intrinsic to the indication. --- ## Team & Environment *(template — fill with real names)* - **Multi-PI structure:** a **phage biology / microbial ecology** PI (cocktail design, host-range matrix, resistance steering) + a **neonatology** PI (NICU surveillance, trial conduct, regulatory). *Why:* the program lives or dies at the lab–clinic interface. - **Key personnel:** strain-resolved metagenomics bioinformatician (Olm/iRep-style analysis); large-animal NEC model expert; GMP/CMC lead; regulatory affairs (phage IND experience); biostatistician. - **Environment:** a NICU with VLBW volume and biobanking; BSL-2 phage facility; preterm-piglet facility (or qualified collaborator); GMP manufacturing access; high-performance computing; pediatric IRB and IND infrastructure. - **Collaborations & open-science commitment:** data/pipeline release plan; collaborator letters for piglet model and GMP manufacturing. --- ## References 1. Olm MR, Bhattacharya N, Crits-Christoph A, et al. *Necrotizing enterocolitis is preceded by increased gut bacterial replication, Klebsiella, and fimbriae-encoding bacteria.* **Science Advances** 5:eaax5727 (2019). 2. Brunse A, Deng L, Pan X, et al. *Fecal filtrate transplantation protects against necrotizing enterocolitis.* **The ISME Journal** 16:686–694 (2022; advance 2021). 3. Spiegelhauer MR, Brunse A, Deng L, et al. *UV-inactivation of the transferred virome abolishes protection against necrotizing enterocolitis* (fecal virome transfer / NEC). **Gut Microbes** (2025). 4. Federici S, Kredo-Russo S, Valdés-Mas R, et al. *Targeted suppression of human IBD-associated gut microbiota commensals by phage combination therapy.* **Cell** 185:2879–2898 (2022). 5. *(Add)* Neu J, Walker WA. *Necrotizing enterocolitis.* **N Engl J Med** (review) — for burden/epidemiology framing. 6. *(Add)* Your institution's NEC epidemiology / cost-of-illness citation. 7. *(Add)* Relevant phage-therapy IND / safety precedent citations for the regulatory section. > ⚠️ **Verify every citation against the primary source before submission** — details (volume/pages/year) are reconstructed here and must be confirmed. Add the bracketed references for a complete bibliography. --- --- # PART B — THE REUSABLE TEMPLATE > Fork this to write *your own* phage grant for a different indication. Replace tokens, then run the prompt library on each section. > **Pre-wired tokens:** `[INDICATION]` · `[POPULATION]` · `[PATHOGEN]` · `[BLOOM/TRIGGER SIGNAL]` · `[ANIMAL MODEL]` · `[KEY PAPER 1–4]` · `[FUNDER]` · `[REVIEW CRITERIA]` · `[MECHANISTIC CONTROL]` --- ## B1. Funder-agnostic proposal skeleton # Title [One precise, benefit-forward title naming PATHOGEN, INDICATION, POPULATION, and the "precision/phage" hook.] # Project Summary / Abstract (~30 lines) - Problem & burden in POPULATION (incidence, mortality, cost, why current care fails) - The mechanistic opening: PATHOGEN bloom / TRIGGER SIGNAL precedes INDICATION [KEY PAPER 1] - Proof that phages are the active protective agent [KEY PAPER 2 + MECHANISTIC CONTROL, KEY PAPER 3] - Translational template: a defined cocktail suppresses PATHOGEN in vivo [KEY PAPER 4] - Central hypothesis (one sentence) - Aims 1–3 in one line each - Impact + open-science commitment # Specific Aims (ONE page) - Problem paragraph → Opportunity (numbered evidence) → Central hypothesis - Aim 1: build/characterize the cocktail + matching engine → Outcome - Aim 2: efficacy + MECHANISTIC CONTROL + PK/PD in ANIMAL MODEL → Outcome - Aim 3: first-in-POPULATION safety/biomarker trial, TRIGGER-based dosing → Outcome - Payoff paragraph # Significance - Burden & unmet need; why supportive care, antibiotics, and non-specific approaches fail - The mechanistic pivot: from "suppress the disease" to "intercept the cause" - Why phage, why now (strain-specific, self-amplifying, non-antibiotic, commensal-sparing) - What changes in the field if you succeed # Innovation - Conceptual inversion (prevention by interception) - Timing/trigger as an engineered active ingredient - A built-in causal control (MECHANISTIC CONTROL) - Resistance pre-emption / phage steering - Commensal-sparing by construction - Open platform # Approach ## Overview & experimental logic (the de-risking staircase) ## Aim 1 — Rationale · Design · Release criteria table · Expected outcomes · Pitfalls & alternatives ## Aim 2 — Rationale · Arms (incl. MECHANISTIC CONTROL) · Endpoints · Power · Expected outcomes · Pitfalls ## Aim 3 — Rationale · Population/trigger · Randomization · Endpoints · Oversight · Outcomes · Pitfalls # Timeline [ILLUSTRATIVE] # Budget Justification Sketch [ILLUSTRATIVE] # Vertebrate Animals (species justification, numbers, welfare, 3Rs) # Human Subjects (population protections, risk rationale, design rigor, inclusion) # Team & Environment (multi-PI lab↔clinic, key personnel, facilities, open-science plan) # References (verify every one against primary source) ``` **Reusable design moves that made Part A strong (steal these):** 1. **Anchor on a pre-symptomatic, measurable trigger** so the intervention is *preventive and timed*, not reactive. 2. **Find your loss-of-function control** (the UV-inactivation move) so one aim proves *mechanism*, not just effect. 3. **Make the cocktail a process, not a product** — ship the matching engine and re-qualification SOP. 4. **Turn externalities into selling points** — commensal sparing vs. the antibiotic it replaces. 5. **Pre-register go/no-go criteria** in tables; reviewers reward decision rules. 6. **Be open** — releasing data/pipelines is both a scientific multiplier and a fundability signal. --- ## B2. The prompt library (adapted from *eseckel/ai-for-grant-writing*) > **Core methodology from the repo:** use AI primarily as a **reviewer and refiner against explicit criteria**, not a ghostwriter. Give it **role + context + the target review criterion**, ask for **specific, actionable feedback**, and **iterate**. Always fact-check AI output — it does not know your data and will invent citations. > > **Prompt-engineering baseline (apply to every prompt below):** (1) assign a role ("act as an NIH study-section reviewer in [INDICATION]"); (2) give context (paste the section + the funder's review criteria); (3) be specific about the output you want; (4) ask for critique + a revised version + the reasoning; (5) iterate. ### 1 — Clarity & lay accessibility - "Act as a NICU clinician with no phage-biology background. Identify every sentence in this [SECTION] that a non-specialist would stumble on, and suggest plainer phrasing without losing precision." - "As a non-native-English reviewer, revise the following for clarity and flow; flag any ambiguous antecedents or run-ons." ### 2 — Compelling narrative / the hook - "Suggest five stronger opening sentences for my Significance that convey the lethality and unmet need of [INDICATION] in [POPULATION] in one breath." - "Review my Abstract's first three sentences. Are they a *hook* or a *throat-clear*? Rewrite for urgency while staying accurate." ### 3 — Structure & flow - "I want to improve the structure of my Specific Aims for [INDICATION]. Propose a reordering that builds a logical de-risking staircase from cocktail → animal mechanism → first-in-human, and explain the logic." - "Critique the flow of my Approach. Does each Aim clearly hand off to the next? Where does the argument jump?" ### 4 — Funder & mission alignment - "Here is [FUNDER]'s mission and this funding opportunity's goals: [PASTE]. Show me, line by line, where my Significance does and does *not* speak to them, and suggest edits to close the gaps." - "Rewrite my closing paragraph to align explicitly with [FUNDER]'s stated priority of [PRIORITY]." ### 5 — Review-criteria alignment (the repo's highest-value move) - "Score my Approach against each NIH criterion — Significance, Innovation, Approach, Investigators, Environment — as a study-section reviewer would (1–9), justify each score, and list the top three fixes that would most raise the overall impact score." - "Here is the specific review criterion: [PASTE]. Give blunt feedback on how well I address it and exactly what to add." ### 6 — Title generation - "Suggest five titles for a grant on a phage cocktail targeting [PATHOGEN] to prevent [INDICATION] in [POPULATION] — each must signal *precision*, *prevention*, and *mechanism*, and stay under 150 characters." ### 7 — Challenge / pitfall identification (pre-empt the reviewer) - "Act as a skeptical study-section reviewer. List the ten most likely objections to this phage-prevention proposal for [INDICATION] — scientific, regulatory, manufacturing, and ethical — and for each, draft a one-paragraph rebuttal or a 'Pitfalls & Alternatives' entry." - "What resistance, PK, or safety concerns will reviewers raise about a live phage biologic in [POPULATION], and how should I address each in the Approach?" ### 8 — Timeline & feasibility - "Given these activities [PASTE], build a feasible 5-year Gantt-style timeline with milestones and go/no-go decision points, front-loading CMC/IND-enabling work." ### 9 — Aim-specific design critique - "Critique my [ANIMAL MODEL] efficacy design. Is my mechanistic control ([MECHANISTIC CONTROL]) sufficient to isolate the active agent? What additional arm or readout would a tough reviewer demand?" - "Review my phase 1b safety design in [POPULATION]: are my stopping rules, DSMB plan, and trigger-based enrollment adequate? What would the FDA flag at pre-IND?" ### 10 — Budget & justification framing - "Given these aims, list the personnel and core cost categories a reviewer expects to see justified, and draft one-sentence justifications tying each to a specific aim." ### 11 — Reverse red-team (run last, before submission) - "You are the assigned reviewer who wants to *triage* this application. Make the strongest case for why it should not be discussed, citing specific weaknesses. Then tell me the minimum set of changes that would move it to 'discuss.'" ### ⚠️ AI guardrails (non-negotiable) - **Never** let AI generate citations — it fabricates them. Verify every reference against the primary literature. - **Never** paste unpublished data, patient information, or confidential collaborator material into a third-party model without authorization. - Treat AI output as a **first-draft critique**, not truth. You own every claim, number, and citation that goes to the funder. - Disclose AI assistance per your funder's and institution's current policy. --- ## B3. One-page "fill-in" worksheet ``` INDICATION: ____________________________________ POPULATION: ____________________________________ PATHOGEN / target: ____________________________________ PRE-SYMPTOMATIC TRIGGER / biomarker: ______________________ KEY PAPER 1 (trigger precedes disease): __________________ KEY PAPER 2 (phage/virome protects): _________________ KEY PAPER 3 (loss-of-function control): _________________ KEY PAPER 4 (defined cocktail in vivo): _________________ ANIMAL MODEL: ____________________________________ MECHANISTIC CONTROL (your "UV-inactivation"): ____________ FUNDER + FOA #: ____________________________________ REVIEW CRITERIA to target: _______________________________ CENTRAL HYPOTHESIS (one sentence): _______________________ GO/NO-GO at end of Aim 1: _________________________________ GO/NO-GO at end of Aim 2: _________________________________ PRIMARY ENDPOINT, Aim 3: __________________________________ OPEN-SCIENCE DELIVERABLES: ________________________________ ``` --- ## License & contribution **CC0 — public domain.** No rights reserved. Fork, adapt, submit, profit, save lives. If this helped you land funding, consider opening *your* funded proposal too — that's how the field compounds. *Built as a public, forkable resource. Methodology adapted from [eseckel/ai-for-grant-writing](https://github.com/eseckel/ai-for-grant-writing). Science anchored on Olm 2019 (Sci Adv), Brunse 2021/22 (ISME J), Spiegelhauer 2025 (Gut Microbes), and Federici 2022 (Cell) — verify all before use.*