Structure characterization with NMR molecular networking

Nuclear Magnetic Resonance (NMR) is among the most widely used techniques for structure determination, yet automated workflows remain underdeveloped compared to mass spectrometry. In this work, we introduce NMR molecular networking and apply it to Heteronuclear Single Quantum Coherence (HSQC) spectra, a key 2D-NMR experiment for structure elucidation. We adapt core principles of MS² networking such as transitivity across multiple spectra, dereplication, and annotation propagation to NMR-driven workflows. First, we develop a modified Hungarian distance metric for HSQC peak matching. Benchmarks show that using this metric, traditional spectral lookup with this score recovers ~70-80% of available structural similarity, but efficiency does not improve when increasing the size of the spectral library. Second, we establish NMR molecular networking using HSQC spectra to propagate annotations and dereplicate compounds. Case studies of experimental natural product spectra demonstrate that annotation transitivity within networks accelerates and improves identification of unknowns. Third, we introduce algorithmic molecular networking, which integrates graph topology metrics to correct inefficient rankings and reduce false positives. Together, these approaches define the first generalizable framework for NMR molecular networking, enabling scalable, high-throughput annotation for natural product discovery and drug development.