SIEVE.X: Structural Identification and Extraction Via Embedded graphs

SIEVE: Structural Identification and Extraction Via Embedded graphs C. Picarelli, G. Raffaini, M. Tommasini Dipartimento di Chimica, Materiali e Ingegneria Chimica “G. Natta”, Politecnico di Milano, Piazza Leonardo da Vinci 32 – 20133 Milano (Italy) email: [email protected] Multiscale modeling is crucial for bridging atomistic processes with macroscopic behavior. SIEVE is an open-source C++ tool developed to automate the extraction of meaningful microstates from molecular dynamics (MD) simulations. Using graph theory, SIEVE reconstructs molecular connectivity from Cartesian coordinates and atomic numbers, enabling fast identification, classification, and indexing of molecular species in trajectories with tens of thousands of atoms. This approach minimizes memory usage, accelerates analysis, and removes manual bottlenecks. The core of SIEVE involves decomposing the simulation box into disjoint molecular graphs, performing graph isomorphism checks against reference structures, and handling efficient I/O through the high-performance H5MD format. Its capabilities directly apply to solvated systems, molecular aggregation, and supramolecular structures. SIEVE provides automated detection and characterization of aggregate size, shape, solvation dynamics, and confinement, ensuring statistically robust analyses. Three representative applications demonstrate its effectiveness: extraction of solvated microstates from MD trajectories of light-emitting compounds for TD-DFT simulations of emission mechanisms; detection of small-molecule confinement within host structures such as nanotubes or fullerenes; and extraction of solvated species for UV–Vis absorption simulations, rationalizing correlations between aggregation, solvation, and conformational effects. SIEVE thus emerges as a transformative tool for multiscale modeling workflows.