Dear-DIAsupXMBD sup

Dear-DIAXMBD analyzes data-independent acquisition (DIA) mass spectrometry proteomics data to identify peptides and proteins without requiring data-dependent acquisition (DDA) spectral libraries.

Key Features:

• Deep Variational Autoencoder Representation Learning: Applies a deep variational autoencoder with triplet loss to learn representations from extracted fragment ion chromatograms.
• Fragment Clustering: Uses k-means clustering to group fragment ions into similar classes based on learned representations.
• Precursor–Fragment Mapping: Constructs inverted index tables to establish relationships between precursor ions, fragment ions, and peptides.
• Spectrum-Centric DIA Analysis: Performs peptide and protein identification from DIA data without relying on spectral libraries derived from data-dependent acquisition (DDA).

Scientific Applications:

• Untargeted Proteomics: Enables peptide and protein identification from complex DIA mass spectrometry datasets.
• Cross-Species Proteomics Studies: Supports analysis of DIA datasets derived from multiple species.
• Mass Spectrometry Data Analysis: Facilitates interpretation of DIA proteomics data generated from different instrument platforms.

Methodology:

Dear-DIAXMBD extracts fragment ion chromatograms from DIA mass spectrometry data, learns feature representations using a deep variational autoencoder with triplet loss, clusters fragments with k-means, and constructs inverted index tables linking precursor ions, fragment ions, and peptides for protein identification.