Computational chemistry represents molecular structures using atomic coordinates without connectivity information. Molecular graphs provide a way to programmatically compare connectivity across different structures and along potential energy surfaces. We present a graph-building approach that is robust across chemical space, enabling the construction of chemically interpretable molecular graphs for unusual bonding or valence structures, including transition metal complexes. With flexible bonding thresholds, this method supports the generation of molecular graphs for transition states and along vibrational trajectories. This enables automated characterisation of transition states in internal coordinates and supports high-throughput exploration of reaction pathways.
Dr. Alister Goodfellow