High-Fidelity Ring Fragments for Molecular Design and Spectroscopy: the PCS-LCB25-Nano-LEGO Framework

Fragment-based molecular design relies on accurate and transferable geometries of modular building blocks, particularly ring systems that are pervasive in medicinal chemistry. Here we present an integrated framework combining the Pisa Composite Schemes (PCS), the LCB25 reference library, and the Nano-LEGO automation tool. In this approach, the most accurate PCS variant delivers benchmark-quality reference structures, while lower-cost density functional theory levels improved by bond corrections serve as validated working models within an automated workflow. LCB25 builds on a previous compilation of accurate fragments by providing over 300 spectroscopically precise equilibrium geometries together with an interactive interface featuring advanced search and retrieval functions. The data set incorporates semiexperimental reference structures, synthon-selected molecules from large quantum-chemical data sets, and all stable nucleobase tautomers together with selected thio- and fluoro-derivatives; in the present work, we further expand it by 44 pharmacologically relevant heterocycles containing nitrogen, oxygen, or sulfur, including selected four- and seven-membered rings. To ensure practical use and on-demand growth, we introduce Nano-LEGO, a fragment-aware workflow that, starting from SMILES strings or rough Cartesian inputs, decomposes targets, checks coverage against the library, generates missing fragments through a robust multilevel optimization pipeline, validates them against benchmark data, and adaptively selects the most suitable whole-molecule working level. An optional conformational search is available for flexible systems. Five drug-like case studies illustrate end-to-end automation, highlighting the gains achieved by bond-corrected protocols at the same cost as standard DFT. Together, PCS, LCB25, and Nano-LEGO provide a modular, high-accuracy infrastructure for fragment-based design, spectroscopy, and data-driven modeling. Automated reconstruction is currently guaranteed for cyclic coverage, with extension to general noncyclic linkers in progress.