Compact SERS detection system enabling automated assay on disc combined with advanced data analysis – A case study of methotrexate

The demand for a compact, all-in-one, sample-to-answer solution for analyzing small molecules within complex matrices is persistent and challenging in analytical chemistry. This manuscript presents a breakthrough in enabling surface-enhanced Raman scattering (SERS) - based analytical assay in a compact detection system, integrating the sample preparation, detection, and data analysis enhanced by machine learning techniques. As a case study, methotrexate (MTX), an anticancer drug, was selected for detection in human serum. The compact tabletop Raman detection unit is equipped with a spectrometer and a motor for the rotation of the incorporated Lab-on-a-Disc unit, which allows simultaneous sample pretreatment of up to eight samples in less than 5 minutes and SERS mapping with a speed of 5 min/analysis. The automated program of the SERS on Disc system simplifies the workflow and enhances the efficiency of analyses. To address the complexity of serum before high sensitivity and specificity detection with SERS, we developed a sample preparation method on a microfluidic disc cartridge that involves precipitation and a nanopillar-assisted separation process, ensuring the isolation of MTX from other analytes. In addition, we implemented a machine learning technique, Partial Least Squares Regression (PLSR), to utilize the fingerprint of SERS spectra for precise and robust quantification of MTX, even in low concentrations. We calculated all figures of merit for the developed method and found that for the PLSR, the accuracy, precision, systematic error, sensitivity, analytical sensitivity, and inverse of analytical sensitivity were 9.41 μM, 9,45 μM, 0.98 μM, 211.64, 1.06 μM⁻¹, and 0.94 μM respectively.