Diagnosis of Early opioids addiction by label-free serum SERS fingerprints with machine learning

Opioids addiction has been steadily rising worldwide, posing serious medical and social challenges. However, there is currently no gold standard for diagnosing opioid addiction for the lack of a specific biomarker. Surface-enhanced Raman scattering (SERS) is widely employed in screening complex components for its capacity to present fingerprint features in the spectra. However, it remains challenging to test blood samples to diagnose diseases without biomarkers using SERS. Here, we introduce a machine learning (ML) assisted, SERS-based platform for analysis of serum of addiction mouse, enabling the diagnosis of morphine and fentanyl addiction. We identify that the proper ligand choice during preparation of the SERS substrate is essential for obtaining clear signal from blood samples. Among several capping ligands, only sodium citrate leads to successful SERS substrate fabrication being able to test blood samples. The critical role of the citrate ligand to promote adsorption and charge transfer between the substrate and the biomolecule of interest is demonstrated. Accuracy rates up to 100% and 87% are achieved for morphine addiction and fentanyl addiction analyses. The present work paves a new way achieving clinical label-free serum diagnosis of morphine and fentanyl addiction by SERS with machine learning, showing its potential for fast-screening of diseases without a clear biomarker.