Virtual Homonuclear Decoupling in Direct Detection NMR Experiments using Deep Neural Networks

Nuclear magnetic resonance (NMR) experiments are frequently complicated by the presence of homonuclear scalar couplings. For the growing body of biomolecular 13C-detected methods, one-bond 13C-13C couplings significantly reduce sensitivity and resolution. The solution to this problem has typically been to record in-phase and anti-phase (IPAP) or spin state selective excitation (S3E) spectra and take linear combinations to yield singlet resolved resonances. This however, results in a doubling of the effective phase cycle and requires additional delays and pulses to create the necessary magnetisation. Here, we propose an alternative method of virtual decoupling using deep neural networks. This methodology requires only the in-phase spectra, halving the experimental time and, by decoupling signals, gives a significant boost in resolution while concomitantly doubling sensitivity relative to the in-phase spectrum. We successfully apply this methodology to virtually decouple in-phase CON (13CO-15N) protein NMR spectra and 13C-13C correlation spectra of protein side chains.