A Transformer-Based Framework for Quality Control of Peptide Tandem Mass Spectra

Area of Science:

• Proteomics
• Computational Biology
• Mass Spectrometry

Background:

• Tandem mass spectrometry (MS/MS) proteomics faces challenges with unidentified spectra due to poor quality, impacting computational efficiency and accuracy.
• Current quality assessment methods lack robustness and generalizability across different instruments.
• Existing deep learning models struggle with global spectral dependencies and data heterogeneity.

Purpose of the Study:

• To develop an automated MS/MS quality assessment framework using a pretrained transformer model.
• To enhance the capture of global spectral relationships and improve cross-instrument transferability.
• To overcome the limitations of traditional and current deep learning approaches in spectral quality assessment.

Main Methods:

• Utilized a transformer framework with self-attention mechanisms for automated MS/MS quality assessment.
• Employed self-supervised pretraining to learn robust, contextualized spectral representations.
• Evaluated the model's performance against existing methods like SPEQ on diverse datasets.

Main Results:

• The transformer framework demonstrated superior performance in automated MS/MS quality assessment.
• The model achieved higher accuracy and enhanced generalization across diverse datasets compared to SPEQ.
• Self-attention effectively captured global fragment relationships, improving spectral identification.

Conclusions:

• The pretrained transformer framework offers a robust and transferable solution for MS/MS spectral quality assessment.
• This approach significantly improves the efficiency and accuracy of proteomics data analysis.
• The model provides a promising advancement for handling challenging spectra in MS/MS-based proteomics.