Multimode Optical Sensor Array Combined with Machine Learning for Accurate Identification and Concentration Prediction of Amino Acid Enantiomers

Simultaneous identification and quantification of multiple amino acid (AA) enantiomers remain as significant challenges. In this study, a multimode, six-channel optical sensor array was developed by three fluorescent/colorimetric carbon dots, which can provide distinctive fluorescence and UV-vis signals to d/l-glutamine, d/l-tryptophan, and d/l-glutamic acid. With the help of machine learning algorithms, this sensor array enabled the accurate qualitative identification and concentration prediction of the three AA enantiomers. The discrimination accuracy for binary and ternary mixtures achieves 95.8%, with an average error of concentration prediction below 10.1%. More importantly, the recognition and prediction performance of this sensor array have been validated using compound amino acid injection and honey samples. A visual assay can also be achieved by the RGB values of the dual-mode sensor array image and SK model, which provide a portable solution for on-site, high-throughput, and naked-eye detection of AA enantiomers.