How NovaMechanics built an automated web tool for constructing Ag, CuO, and TiO₂ spherical nanoparticles in silico and calculating their atomistic descriptors — bridging digital nanostructure design with machine learning-ready data. Bridging the Gap Between NP Synthesis and Computational Modelling Ag, CuO, and TiO₂ nanoparticles are among the most widely produced nanomaterials, used in heating/cooling systems, electronics, medical devices, and nanocomposites. Yet comprehensive risk assessments of their potential hazards remain limited due to the high cost and complexity of experimental testing. Existing QSAR/QPAR models for NP toxicity prediction are constrained by dataset size and lack universally accepted structural descriptors. A computational tool was needed that could digitally construct NPs and automatically extract atomistic descriptors to serve as inputs for machine learning models — enabling in silico safety assessment before synthesis. 4 Material types supported (Ag, CuO, TiO₂-Anatase, TiO₂-Rutile) 30+ Atomistic descriptors calculated per NP Zero Code: No programming needed — fully automated backend Select crystallographic information files: Pre-selected CIF files from the Crystallography Open Database for each material type (Ag: Fm-3m space group, CuO: C12/c1, TiO₂-Anatase: I41/amd, TiO₂-Rutile: P4₂/mnm). The tool provides default values suitable for non-expert users while allowing advanced customisation. Geometrically construct spherical NPs: Replicate the unit cell to create a bounding box, remove atoms outside the target sphere diameter, and enforce stoichiometric correctness using an iterative shell-removal algorithm (0.02 Å shell thickness) to ensure electrically neutral, realistic NP structures. Energy minimization via LAMMPS: Apply molecular dynamics energy minimization using the LAMMPS simulator with force fields sourced from the OpenKIM database (EAM, MEAM, Buckingham, COMB3). Reactive force fields enable bond breaking and formation at NP surfaces for realistic structures. Calculate atomistic descriptors: Automatically compute 30+ descriptors including average potential energy per atom, coordination numbers, common neighbour parameters, and hexatic order parameters — separately for core (>4 Å from surface) and shell (≤4 Å) regions of each NP. Upload to NanoPharos database: Enable direct upload of constructed digital NPs and their descriptors to the NanoPharos database, facilitating FAIR data sharing and reuse in QSAR model development across the nanoinformatics community. 4 Material Types: Ag, CuO, TiO₂-Anatase, and TiO₂-Rutile fully supported Automated End-to-End Pipeline: From material selection to ML-ready descriptors without coding 30+ Atomistic Descriptors: Core/shell separation for energy, coordination, and order parameters LAMMPS MD Integration: Energy minimization via LAMMPS with multiple OpenKIM force fields FAIR Data Upload: Direct NanoPharos database integration for data sharing and reuse Free Web Access: Freely available on the Enalos Cloud Platform
