Machine Learning-Empowered FTIR Spectroscopy Serum Analysis Stratifies Healthy, Allergic, and SIT-Treated Mice and Humans

Abstract

The unabated global increase of allergic patients leads to an unmet need for rapid and inexpensive tools for the diagnosis of allergies and for monitoring the outcome of allergen-specific immunotherapy (SIT). In this proof-of-concept study, we investigated the potential of Fourier-Transform Infrared (FTIR) spectroscopy, a high-resolution and cost-efficient biophotonic method with high throughput capacities, to detect characteristic alterations in serum samples of healthy, allergic, and SIT-treated mice and humans. To this end, we used experimental models of ovalbumin (OVA)-induced allergic airway inflammation and allergen-specific tolerance induction in BALB/c mice. Serum collected before and at the end of the experiment was subjected to FTIR spectroscopy. As shown by our study, FTIR spectroscopy, combined with deep learning, can discriminate serum from healthy, allergic, and tolerized mice, which correlated with immunological data. Furthermore, to test the suitability of this biophotonic method for clinical diagnostics, serum samples from human patients were analyzed by FTIR spectroscopy. In line with the results from the mouse models, machine learning-assisted FTIR spectroscopy allowed to discriminate sera obtained from healthy, allergic, and SIT-treated humans, thereby demonstrating its potential for rapid diagnosis of allergy and clinical therapeutic monitoring of allergic patients.