Application of a Hopfield neural network to extract essential spectral patterns from ToF-SIMS spectra of peptides

A recurrent neural network (RNN) system based on a Hopfield neural network (HNN) was developed to extract essential spectral patterns from the time-of-flight secondary ion mass spectrometry (ToF-SIMS) spectra of peptide samples. Because ToF-SIMS produces various fragment ions from organic molecules, the interpretation of ToF-SIMS spectra is generally complicated. ToF-SIMS is useful for peptide analysis because it detects specific amino acid fragment ions from peptides that indicate peptide information. However, the ToF-SIMS spectra also contain fragment ions that do not preserve the main structures of the original molecules, which makes them difficult to interpret. Therefore, it is crucial to extract essential spectral patterns from the ToF-SIMS spectra of organic materials. Peptides were selected as the target organic materials for this study due to their systematic chemical structures. A modified HNN was trained on the ToF-SIMS spectra of each peptide, and the trained HNNs were then used to recall patterns for various peptide ToF-SIMS spectra. The results show that the modified HNN recall essential spectra containing specific ions, including the protonated molecular ions and amino acid fragment ions of target peptides. Furthermore, the HNN results revealed differences and similarities between peptides with similar and different amino acid sequences. Thus, this study demonstrates the effectiveness of the HNN in interpreting complex spectra and its potential for preprocessing data for further analysis.