A Novel Peptides Database Approach for Enhanced Dereplication of Peptaibols Using Molecular Network Based on the <i>t</i> -SNE Algorithm

Peptaibols are nonribosomal peptides produced by <i>Trichoderma</i> species and are rich in nonproteinogenic amino acids such as α-aminoisobutyric acid (Aib). Their pronounced structural diversity, and underrepresentation in spectral libraries pose significant challenges for dereplication in high-throughput liquid chromatography coupled with tandem high resolution mass spectrometry (LC-HRMS/MS) workflows. This study introduces a novel strategy that integrates theoretical MS/MS fragmentation prediction, intensity modeling based on experimental data, and dimensionality reduction via <i>t</i>-distributed stochastic neighbor embedding (<i>t</i>-SNE) to improve peptaibol dereplication and classification. A custom spectral database was generated using predicted b- and y-type ions, with fragment intensities calibrated to match observed fragmentation patterns. The method demonstrated strong robustness across a diverse range of peptaibols, allowing clearer sequence assignments. One limitation was noted with Alamethicin F-50, where missing sequence motifs in the intensity model slightly reduced specificity. Nonetheless, the approach supports fast and reliable classification of unknown peptaibols through fragment-based matching, offering a powerful tool for prioritizing bioactive molecular families. To our knowledge, this is the first implementation of a <i>t</i>-SNE-guided theoretical spectral database for dereplication and classification of peptide-like natural products. Ongoing enrichment of the database is expected to further expand its specificity and applicability across broader peptides families.