DeepSecMS Advances DIA-Based Selenoproteome Profiling Through Cys-to-Sec Proxy Training.

Area of Science:

• Biochemistry
• Proteomics
• Bioinformatics

Background:

• Selenoproteins, rare proteins containing selenocysteine (Sec), are crucial for various biological functions.
• Only 25 human selenoproteins are known, necessitating advanced methods for comprehensive selenoproteome analysis.
• Existing tools like SecMS and SIS offer potential for discovering new selenoproteins.

Purpose of the Study:

• To develop a deep learning method, DeepSecMS, for accurate selenoprotein identification.
• To enhance the discovery of novel selenoproteins and rare post-translational modifications.
• To improve the comprehensive analysis of selenoproteomes.

Main Methods:

• Developed DeepSecMS, a deep learning approach for selenoprotein identification.
• Utilized a proxy training strategy with cysteine-containing peptides to generate theoretical selenocysteine-containing peptides.
• Integrated DeepSecMS with data-independent acquisition (DIA) mass spectrometry.

Main Results:

• DeepSecMS accurately predicts key features of Sec-containing peptides, including MS2 spectra, retention time (RT), and ion mobility (IM).
• Integration with DIA methods significantly improved the identification of known selenoproteins across various tissues.
• Identified numerous high-scoring potential novel selenoproteins, expanding the known human selenoproteome.

Conclusions:

• DeepSecMS is a powerful tool for advancing selenoprotein research and discovering novel selenoproteins.
• The proxy training strategy shows promise for analyzing other rare post-translational modifications.
• This work significantly enhances the ability to comprehensively analyze selenoproteomes.