DeepPhospho accelerates DIA phosphoproteome profiling through in silico library generation.

Area of Science:

• Proteomics
• Bioinformatics
• Systems Biology

Background:

• Data-independent acquisition (DIA) phosphoproteomics offers superior quantification but relies on data-dependent acquisition (DDA) spectral libraries.
• DDA library construction is a bottleneck, limiting throughput, proteome coverage, and increasing sample requirements for DIA phosphoproteomics.

Purpose of the Study:

• To introduce DeepPhospho, a deep neural network for predicting LC-MS/MS data of phosphopeptides.
• To establish a DDA-independent DIA phosphoproteomics workflow using in silico libraries generated by DeepPhospho.
• To enhance phosphoproteome coverage and discovery of signaling pathways and kinases.

Main Methods:

• Development of DeepPhospho, a deep neural network for phosphopeptide spectral data prediction.
• Generation of in silico spectral libraries using DeepPhospho.
• Implementation of a DIA phosphoproteomics workflow utilizing DeepPhospho predicted libraries for data mining.
• Comparative analysis with DDA library-based DIA phosphoproteomics.

Main Results:

• DeepPhospho enables accurate prediction of LC-MS/MS data for phosphopeptides.
• The DeepPhospho-empowered DIA workflow circumvents the need for DDA library construction.
• This workflow significantly expands phosphoproteome coverage and maintains high quantification performance.
• More signaling pathways and regulated kinases were discovered compared to the DDA library-based approach.

Conclusions:

• DeepPhospho facilitates a DDA-independent DIA phosphoproteomics workflow, improving analytical efficiency.
• The developed workflow enhances the depth and breadth of phosphoproteome profiling.
• DeepPhospho empowers broader discovery of biological signaling networks and kinase activities.