Pre-trained Maldi Transformers improve MALDI-TOF MS-based prediction

Gaetan De Waele, Gerben Menschaert, Peter Vandamme, Willem Waegeman

Abstract: Highlights•We propose a transformer neural network for MALDI-TOF MS data.•We develop the first pre-training approach specifically adapted for MS data.•We test our models using an extensive set of datasets and baseline methods.•We show that our pre-trained model can also be used to denoise MALDI-TOF spectra.