Simultaneous classification of multiple classes in NMR metabolomics and vibrational spectroscopy using interval-based...

Area of Science:

• Chemometrics
• Spectroscopy
• Data Analysis

Background:

• Interval-based chemometric algorithms offer powerful solutions for spectral alignment, regression, and classification.
• These methods enhance model performance, reduce complexity, and improve spectral interpretation.

Purpose of the Study:

• To introduce and evaluate the interval Extended Canonical Variate Analysis (iECVA) method.
• To compare the performance of iECVA with interval Partial Least Squares Discriminant Analysis (iPLS-DA).
• To assess the utility of interval-based methods for enhancing spectral interpretation and biomarker discovery.

Main Methods:

• Development and application of interval Extended Canonical Variate Analysis (iECVA).
• Comparison with interval Partial Least Squares Discriminant Analysis (iPLS-DA).
• Testing on three spectroscopic datasets: Nuclear Magnetic Resonance (NMR), Near Infrared (NIR), and Infrared (IR).

Main Results:

• Interval-based methods significantly enhance model interpretability by identifying crucial spectral regions.
• Both iECVA and iPLS-DA demonstrated similar performance in terms of misclassifications and identified regions.
• Interval Partial Least Squares Discriminant Analysis (iPLS-DA) consistently yielded lower model complexity compared to iECVA.

Conclusions:

• Interval-based classification methods are effective for enhancing interpretability and facilitating biomarker discovery in spectroscopic data.
• iECVA and iPLS-DA offer valuable tools for multivariate data analysis, with iPLS-DA showing advantages in model simplicity.
• Freely available Matlab source codes for iECVA and iPLS-DA are provided to promote further research.