Background. New Delhi metallo-beta-lactamase 1 (NDM-1) hydrolyses almost every clinically useful beta-lactam and is untouched by the serine-beta-lactamase inhibitors in current use. Repurposing an approved drug as an NDM-1 inhibitor would shorten the path to a carbapenem adjuvant, and a substantial computational literature now proposes candidates on the basis of docking and molecular dynamics. Whether that literature has converged on anything has not been examined.

Objective. To characterise the published computational repurposing evidence against NDM-1, and to test whether the compounds it proposes differ chemically from the compounds that have actually been shown to inhibit the enzyme.

Methods. Thirty-two computational studies of NDM-1 and related class B1 metallo-beta-lactamases were retrieved and coded against fixed methodological items. Structures for 30 compounds named in this literature were assembled from DrugBank identifiers, from published chemical infoboxes, or built from the published IUPAC name and accepted only on an exact molecular-formula match. Calculated physicochemical properties and the presence of 13 classes of zinc-coordinating motif were computed in RDKit. Reported potencies were collected in micromolar units and grouped by the route through which each compound was selected. Groups were compared with exact and rank-based tests.

Results. Of 32 studies, 7 (21.9 per cent) state how the two active-site zinc ions were parameterised, 3 (9.4 per cent) report a pose-recovery control, 2 (6.2 per cent) combine more than one scoring function, and 12 (37.5 per cent) report any experimental corroboration. Every study published in the final three years of the corpus is computational only. Fourteen of 16 compounds with measured NDM-1 inhibition carry a recognised zinc-binding motif, against 0 of 4 compounds proposed by docking alone (Fisher exact p = 0.0031; whole-set sensitivity analysis 15 of 17 against 0 of 7, p = 0.0001). Docking-only candidates are heavier than corroborated ones (median calculated molecular weight 576.54 against 288.75 Da, p = 0.0071). Compounds found by structure-based virtual screening have a median reported IC50 of 27.10 micromolar, against 0.04 micromolar for the designed inhibitors now in clinical development.

Conclusion. The NDM-1 repurposing literature is converging on a chemotype the enzyme cannot use. Docking scores computed against a rigid, singly parameterised di-zinc site have selected large lipophilic drugs with no metal-coordinating chemistry, and no approved drug has been independently re-identified across studies. A short list of reporting requirements is proposed.