An intensity-region driven multi-classifier scheme for improving the classification accuracy of proteomic MS-spectra

Area of Science:

• Biomedical Engineering
• Computational Biology
• Analytical Chemistry

Background:

• Mass spectrometry (MS) is crucial for biomarker discovery in diseases like ovarian cancer.
• Accurate classification of MS spectra is essential for early diagnosis and treatment.
• Existing methods may benefit from enhanced feature extraction and classification strategies.

Purpose of the Study:

• To develop and evaluate a pattern recognition system for improved MS-spectra classification.
• To enhance the accuracy of distinguishing normal from ovarian cancer MS-spectra.
• To identify potential biomarkers from MS-spectra intensity regions.

Main Methods:

• A pattern recognition system was designed using ensemble learning with a majority vote combination.
• MS-spectra were automatically segmented into common intensity regions.
• Informative features (m/z values) were extracted from each region.
• Support Vector Machine, Probabilistic Neural Network, and k-Nearest Neighbour classifiers were employed in a multi-classifier scheme.
• The system was validated using external cross-validation on a public ovarian proteomic dataset.

Main Results:

• The system achieved an average overall performance of 97.18% in discriminating normal from ovarian cancer MS-spectra.
• High mean sensitivity (98.52%) and specificity (94.84%) were recorded.
• The ensemble approach effectively integrated information from different MS-spectra intensity regions.

Conclusions:

• The proposed pattern recognition system significantly improves the classification accuracy of ovarian cancer MS-spectra.
• The method demonstrates robust and reliable performance for biomarker discovery and cancer diagnostics.
• This approach offers a promising tool for the clinical application of MS-based proteomic analysis.