Spiro-Terpenoid Privileged Scaffolds: An Auditable Multiscale Workflow for Undruggable Targets

Three-dimensional scaffold design is often judged by saturation proxies that do not distinguish local graph constraint from global conformational preorganization. We developed an auditable, computation-light workflow for a prospective 25-member spiro-terpenoid library directed at KRAS G12D and the MYC G-quadruplex. Matched graph controls were combined with three independently seeded conformer searches, GFN2-xTB single-point energies, a 20,000-draw descriptor-and-weight ensemble, three-seed AutoDock Vina calculations, and explicit-water complex relaxation. The selected candidates, A-03 and B-02, retained near-orthogonal local ring planes across conformer pools, but global shape breadth and finite-pool entropy did not support a universal spiro preorganization advantage. A-03 ranked first for KRAS in 99.9% of ensemble draws and B-02 ranked first for the G-quadruplex in 87.5%. In 200 ps unrestrained production trajectories, target-aligned ligand root-mean-square deviations were 0.84 and 1.27 Å, respectively; geometry-compatible hydrogen bonds occurred in 95% and 10% of saved frames. These short trajectories support only early pose relaxation and contact persistence, not equilibrium stability, kinetics, residence time, or affinity. Complete trajectories, analysis code, and checkpoint-ready long-trajectory systems make every decision layer reproducible and experimentally falsifiable. The workflow separates topological design hypotheses from thermodynamic claims and identifies synthesis priorities without wet-laboratory data.