A Novel Peptides Database Approach for Enhanced Dereplication of Peptaibols Using Molecular Network Based on the t‑SNE Algorithm

Peptaibols are nonribosomal peptides produced by Trichoderma 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 t-distributed stochastic neighbor embedding (t-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 t-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.