Quality assessment of tandem mass spectra using support vector machine (SVM)

Area of Science:

• Proteomics
• Biotechnology
• Analytical Chemistry

Background:

• Tandem mass spectrometry is crucial for identifying proteins in complex mixtures.
• Existing peptide identification methods like SEQUEST and MASCOT rely on comparing experimental spectra with theoretical spectra from databases.
• A significant portion of mass spectrometry data is of insufficient quality for interpretation, necessitating effective filtering strategies.

Purpose of the Study:

• To develop and validate a machine learning-based algorithm for assessing tandem mass spectrometry (MS/MS) spectral quality.
• To improve the efficiency and accuracy of peptide identification by pre-filtering low-quality spectra.

Main Methods:

• A support vector machine (SVM) model was developed to classify MS/MS spectra quality.
• Each spectrum was represented by 16 distinct features.
• Four SVM classifiers were trained and tested using ISB and TOV datasets, leveraging SEQUEST search results.

Main Results:

• The proposed SVM-based approach demonstrated superior performance in spectral quality assessment compared to existing methods.
• Validation using MASCOT search results confirmed the effectiveness of the SVM classifiers.
• The method successfully distinguished between high- and low-quality spectra.

Conclusions:

• The developed SVM method effectively filters poor-quality spectra prior to database searching in proteomics.
• This approach facilitates the identification of more peptides, including post-translational modifications, by optimizing the use of high-quality spectra with various search engines or de novo sequencing.