C07 - Untangling multi-property NMR signals in drug screening with data-driven neural networks

The goal of this project is to tackle the effects of signal coupling in parallel NMR via artificial intelligence (AI). We will develop a transformer-based neural network that takes a set of coupled spectra from parallel NMR as input, and extracts the relevant information from each signal, such that it can output clean, decoupled spectra for each individual sample. We will use self-supervised pre-training on existing data to minimize data collection and annotation, and to investigate how much the model can learn about the underlying physical principles governing nuclear magnetic resonance.

Team