Characterization of Breast Cancer via Intrinsic Fluorescence of Tryptophan and Other Biomolecules using Machine Learning Models

Background/Objectives Breast cancer (BC) remains the most prevalent cancer among women, and its effective treatment relies heavily on early detection. To our knowledge, this is the first study to integrate serum tryptophan fluorescence data with Gaussian Process Regression (GPR) and validate findings through confocal and histopathological imaging. Optical techniques are employed to assess fluorescence variations of key endogenous biomolecules, including tryptophan, Reduced Nicotinamide Adenine Dinucleotide (NADH), and Flavin Adenine Dinucleotide (FAD), which play vital roles in cellular metabolism. Changes in their concentrations are associated with malignancy. The objective is to distinguish between normal and cancerous breast tissues based on tryptophan fluorescence spectra. Methods Intrinsic fluorescence spectra of tryptophan, NADH, and FAD were acquired and compared to assess their potential for distinguishing between normal and cancerous conditions. Among them, tryptophan fluorescence intensity in serum emerged as the most effective marker for differentiation. GPR and other potential models were applied to predict cancer status based on serum fluorescence data, with model performance evaluated using standard error metrics such as root mean squared error (RMSE). Additionally, confocal autofluorescence imaging of breast tissue samples was performed and compared with Hematoxylin and Eosin (H&E)-stained Differential Interference Contrast (DIC) images to validate morphological differences and support the method’s effectiveness. Results The findings reveal a clear distinction between normal and cancerous breast tissues based on serum tryptophan fluorescence intensity. Spectral analysis confirmed that tryptophan, along with NADH and FAD, effectively reflects malignancy-associated metabolic alterations. Among the models tested, GPR outperformed with the highest predictive accuracy, with RMSEs of 1847.5 (range: 2 × 106) for normal and 3621.6 (range: 6 × 106) for cancerous samples, both constituting less than 0.1% of their respective fluorescence intensity ranges. Conclusions Serum tryptophan fluorescence intensity has been identified as a robust biomarker for distinguishing between normal and cancerous breast tissues. The integration of optical spectral data with the GPR model yielded high predictive accuracy, demonstrating the method’s effectiveness in identifying malignancy. The results were supported by stained and confocal imaging. These findings support the development of a non-invasive, real-time screening method that could reduce reliance on traditional histopathological diagnostics and serve as a rapid, cost-effective adjunct to biopsy and mammography in early BC detection.