Novel diazabicyclooctane (DBO) beta-lactamase inhibitor analogues, designed via multi-feature virtual screening, inhibit the KPC-3 carbapenemase at the active site with predicted binding affinities competitive with the clinical DBO inhibitors avibactam and relebactam, providing a next-generation inhibitor series to restore carbapenem efficacy against carbapenem-resistant Enterobacteriaceae (CRE).

At least one of a defined set of N (≤50) novel diazabicyclooctane (DBO) analogues, generated via multi-feature virtual screening (docking score + pharmacophore + ADMET filters), will exhibit: (a) a computationally predicted binding free energy (ΔG, MM-GBSA or FEP) to KPC-3 within ±1.5 kcal/mol of avibactam's reference value under the same scoring protocol, AND (b) upon synthesis and biochemical testing, an experimentally measured carbamoylation rate constant (k2/K or koff) and enzyme inhibition IC50/Ki against purified KPC-3 that is within 1 order of magnitude of avibactam's published values (Ki ~ 10-100 nM range depending on assay conditions), AND (c) restoration of meropenem susceptibility (≥4-fold MIC reduction) in at least one KPC-3-producing Enterobacteriaceae clinical isolate. Failure of any one of (a)-(c) falsifies the specific compound; failure across the full candidate set falsifies the hypothesis as stated.

Minimum viable test (MVT): computational-to-biochemical funnel on a reduced set.

1. Computational validation tier: re-dock avibactam, relebactam, and 3-5 literature DBO analogues with published KPC-3 Ki values into KPC-3 crystal structure (PDB 6D16, 6XFV, or equivalent) to establish scoring-function benchmark correlation (r² target ≥0.5).
2. Virtual screen 500-2,000 novel DBO analogues (enumerated via scaffold decoration of the bicyclic urea/sulfate core) using consistent docking + MM-GBSA rescoring.
3. Select top 10-20 by composite score (docking + ADMET + synthetic accessibility, e.g., SAscore ≤4).
4. Synthesize top 5-8 candidates (medicinal chemistry, parallel synthesis).
5. Biochemical IC50/Ki determination against purified KPC-3 (nitrocefin hydrolysis assay, steady-state kinetics).
6. MIC testing (broth microdilution, CLSI M07) of top 2-3 compounds + meropenem (fixed ratio or checkerboard) against a panel of ≥10 KPC-3-producing clinical isolates.
7. Compare all experimental results back to computational predictions to close the loop and quantify predictive accuracy.

# Stage 1: Computational benchmark & screening
load_structures(pdb_ids=["6D16","6XFV","6V7H"])
benchmark_set = load_known_inhibitors_with_ki()  # n>=15
docking_protocol = configure_glide_or_autodock(scoring="SP+MMGBSA")
validate_protocol(docking_protocol, benchmark_set, rmsd_threshold=2.0, r2_threshold=0.5)
if not validated: ABORT

library = enumerate_dbo_analogues(core_scaffold="bicyclic_urea_sulfate", n=1000-2000)
library = apply_filters(library, admet_model, synth_accessibility_max=4)
scores = dock_and_rescore(library, docking_protocol)
top_candidates = rank_composite(scores, weights={"affinity":0.5,"admet":0.3,"synthesis":0.2})[:20]

# Stage 2: Wet-lab validation
for compound in top_candidates[:8]:
    synthesize(compound)  # medchem, 2-3 step route
    confirm_purity(compound, methods=["LCMS","NMR"], threshold=0.95)

kpc3 = express_and_purify_recombinant_protein()
for compound in synthesized_candidates:
    ic50 = nitrocefin_inhibition_assay(kpc3, compound, replicates=3)
    if ic50 <= 1e-6:
        mechanism = determine_kinetics(kpc3, compound)  # koff, k2/K
        adduct_confirmed = intact_ms(kpc3, compound)

# Stage 3: Phenotypic validation
isolates = load_cdc_ar_isolate_bank(genotype="KPC-3")
for compound in top_candidates_with_confirmed_covalent_binding:
    mic_results = checkerboard_assay(compound, meropenem, isolates)
    report(mic_results, fold_reduction_threshold=4)

# Stage 4: Closure — compare predicted vs observed
correlation_report(predicted_affinities, observed_ic50s)