Application of SERS method in combination with machine learning for therapeutic drug monitoring data.

The proposed approach is suitable for further validation and potential implementation in clinical practice to access the safety of methotrexate therapy in patients receiving this drug in low-dose regimen. SERS spectra were successfully recorded from serum samples obtained from patients after methotrexate administration and from healthy volunteers who had not received methotrexate.

Flat silver nanotriangles deposited on quartz glass were used as a SERS-active substrate to enhance the Raman signal and enable the detection of molecular changes in serum samples. The obtained spectra were processed in three ways: without spectra processing, processing with a baseline correction, processing with a baseline and normalization. The ability to classify SERS spectra from patients with MTX concentrations above and below the cutoff concentration of 0.71 μM using Partial Least Squares (PLS) and Discriminant Analysis (DA) methods was demonstrated.</p>