Abstract
<jats:p>High-throughput computational screening of metal–organic frameworks (MOFs) for gas storage and separation applications routinely begins from structural repositories containing tens of thousands of candidates, only a small fraction of which are geometrically suitable for further simulation. We report a geometry-based pre-screening dataset of 12,597 lattice-verified MOF structures, filtered from an initial pool of approximately 71,500 candidates drawn from four public repositories (pore-limiting diameter ≥ 3.3 Å, crystal density ≥ 0.1 g cm-3, and related structural criteria). Rather than ranking structures by simple maximization of pore size or minimization of density– an approach that disproportionately favors extreme, often mechanically unstable frameworks – each structure is scored against a trapezoidal target-range function that rewards geometric properties falling within ranges considered reasonable for stable, practically usable MOFs. The dataset, its scoring methodology, and an independent reproducibility check (a standalone script that recomputes all scores from raw geometric columns and matches the published values to floating-point precision) are made available to support downstream grand canonical Monte Carlo (GCMC) adsorption screening campaigns. Two methodological limitations identified during quality control – a large tied-score plateau affecting 42.6% of structures, and duplicate entries arising from alternate relaxation protocols – are documented and addressed with a deduplicated shortlist.</jats:p>