Boosting MS1-only Proteomics with Machine Learning Allows 2000 Protein Identifications in Single-Shot Human Proteome...

Area of Science:

• Proteomics
• Mass Spectrometry
• Biomedical Research

Background:

• Proteome-wide analyses using tandem mass spectrometry are time-consuming, hindering clinical applications.
• DirectMS1 offers a faster alternative using ultrashort LC gradients and MS1-only spectra acquisition.
• Current DirectMS1 identifies ~1000 proteins in minutes at 1% FDR.

Purpose of the Study:

• To enhance the DirectMS1 method's protein identification capabilities.
• To integrate advanced machine learning and retention time prediction for improved performance.
• To explore additional techniques like FAIMS for further optimization.

Main Methods:

• Applied LightGBM decision tree boosting algorithm for peptide feature matching in MS1 spectra.
• Integrated DirectMS1 with DeepLC for accurate peptide retention time prediction.
• Utilized Liquid Chromatography-Field Asymmetric Ion Mobility Spectrometry-MS1 (LC-FAIMS/MS1) analysis.

Main Results:

• Achieved identification of over 2000 proteins at 1% FDR from HeLa cell line.
• Reduced analysis time to 5 minutes using the enhanced DirectMS1 method.
• Demonstrated improved performance through machine learning and FAIMS integration.

Conclusions:

• The enhanced DirectMS1 method significantly increases proteome coverage and speed.
• Machine learning and FAIMS integration are effective strategies for accelerating proteomic analyses.
• This advancement holds promise for broader adoption of proteomics in clinical and biomedical research.