Molecular profiling of blood plasma-derived extracellular vesicles derived from Duchenne muscular dystrophy patients through integration of FTIR spectroscopy and machine learning reveals disease signatures

Purpose: To identify and monitor the FTIR spectral signatures of plasma extracellular vesicles (EVs) from Duchenne Muscular Dystrophy (DMD) patients at different stages with Healthy controls using machine learning models.

Materials and methods: Whole blood samples were collected from the DMD (n = 30) and Healthy controls (n = 12). EVs were extracted by the Total Exosome Isolation (TEI) Method and resuspended in 1XPBS. We characterize the morphology, size, particle count, and surface markers (CD9, Alix, and Flotillin) by HR-TEM, NTA, and Western Blot analysis. The mid-IR spectra were recorded from (4000-400 cm^-1) by Bruker ALPHA II FTIR spectrometer model, which was equipped with an attenuated total reflection (ATR) module. Machine learning algorithms like Principal Component Analysis (PCA) and Random Forest (RF) for dimensionality reduction and classifying the two study groups based on the FTIR spectra. The model performance was evaluated by a confusion matrix and the sensitivity, specificity, and Receiver Operating Characteristic Curve (ROC) was calculated respectively.

Results: Alterations in Amide I & II (1700-1470 cm^-1) and lipid (3000-2800 cm^-1) regions in FTIR spectra of DMD compared with healthy controls. The PCA-RF model classified correctly the two study groups in the range of 4000-400 cm^-1 with a sensitivity of 20 %, specificity of 87.50 %, accuracy of 71.43 %, precision of 33.33 %, and 5-fold cross-validation accuracy of 82 %. We analyzed the ten different spectral regions which showed statistically significant at P < 0.01 except the Ester Acyl Chain region.

Conclusion: Our proof-of-concept study demonstrated distinct infrared (IR) spectral signatures in plasma EVs derived from DMD. Consistent alterations in protein and lipid configurations were identified using a PCA-RF model, even with a small clinical dataset. This minimally invasive liquid biopsy method, combined with automated analysis, warrants further investigation for its potential in early diagnosis and monitoring of disease progression in DMD patients within clinical settings.