DeepPhospho

DeepPhospho predicts fragment ion intensities and indexed retention times (iRT) for phosphopeptides using deep learning to generate in silico libraries for data-independent acquisition (DIA) phosphoproteome profiling.

Key Features:

• Hybrid Neural Network Architecture: A hybrid deep neural network integrating Long Short-Term Memory (LSTM) networks with transformer modules predicts fragment ion intensity and indexed retention time (iRT) for phosphopeptides.
• Spectral Library Generation: Generates in silico spectral libraries for DIA, eliminating the need for project-specific data-dependent acquisition (DDA) spectral library construction.
• Improved Phosphoproteome Coverage: Extends phosphoproteome coverage compared with DDA-based library workflows, enabling identification of more signaling pathways and regulated kinases as demonstrated in EGF signaling studies.
• High Quantification Performance: Maintains high quantification reproducibility and accuracy for DIA phosphoproteomics data mining.

Scientific Applications:

• Cellular signaling analysis: Enables deeper profiling of phosphorylation events to map signaling pathway activity.
• Disease mechanism investigation: Supports characterization of phosphorylation changes relevant to physiological and pathological processes.
• Biomarker discovery: Facilitates identification of phosphorylation-based biomarker candidates from DIA phosphoproteomics datasets.
• Identification of regulatory proteins: Aids detection of regulated kinases and other signaling proteins involved in cellular regulation.

Methodology:

Hybrid deep neural network combining LSTM and transformer modules predicts fragment ion intensities and iRT for phosphopeptides and generates in silico spectral libraries for DIA phosphoproteomics.