Spectroscopic and machine learning approaches for clinical subtyping in systemic sclerosis

Area of Science:

• Biomedical Spectroscopy
• Immunodermatology
• Computational Biology

Background:

• Systemic sclerosis (SSc) is a complex autoimmune disorder with diverse clinical manifestations.
• Current diagnostic and stratification methods for SSc can be invasive and time-consuming.
• Identifying non-invasive tools for early disease detection and subtype classification is crucial.

Purpose of the Study:

• To investigate the utility of Fourier-transform infrared (FTIR) spectroscopy on whole blood samples for SSc classification.
• To explore the application of multivariate and machine learning techniques for SSc subtype differentiation.
• To assess the potential of FTIR spectroscopy as a non-invasive tool for SSc biomarker discovery.

Main Methods:

• Whole blood samples from SSc patients were analyzed using FTIR spectroscopy.
• Multivariate analysis, including Principal Component Analysis (PCA), was employed to analyze spectral data.
• Supervised machine learning models, such as Random Forest (RF), were developed for classification tasks.

Main Results:

• FTIR spectroscopy revealed subtle but consistent spectral differences in amide I/II and lipid-associated regions.
• PCA demonstrated clear clustering of samples, indicating distinct spectral profiles.
• The Random Forest model achieved optimal performance in classifying diffuse versus limited SSc subtypes.

Conclusions:

• FTIR spectroscopy combined with machine learning shows promise as a non-invasive method for SSc disease stratification.
• This approach has the potential for biomarker discovery in systemic sclerosis.
• Further optimization of models and spectral feature extraction is necessary for clinical implementation.