Network Analysis Reveals Spatial Clustering and Annotation of Complex Chemical Spaces: Application to Astrochemistry

Abstract

How are molecules linked to each other in complex systems? In a proof-of-concept study, we have developed the method mol2net (https://zenodo.org/record/7025094) to generate and analyze the molecular network of complex astrochemical data (from high-resolution Orbitrap MS1 analysis of H2O:CH3OH:NH3 interstellar ice analogs) in a data-driven and unsupervised manner, without any prior knowledge about chemical reactions. The molecular network is clustered according to the initial NH3 content and unlocked HCN, NH3, and H2O as spatially resolved key transformations. In comparison with the PubChem database, four subsets were annotated: (i) saturated C-backbone molecules without N, (ii) saturated N-backbone molecules, (iii) unsaturated C-backbone molecules without N, and (iv) unsaturated N-backbone molecules. These findings were validated with previous results (e.g., identifying the two major graph components as previously described N-poor and N-rich molecular groups) but with additional information about subclustering, key transformations, and molecular structures, and thus, the structural characterization of large complex organic molecules in interstellar ice analogs has been significantly refined.