Multi-way calibration strategies involving excitation-emission data measurements from drug-modulated fluorescence in CdTe quantum dots and AgInS nanocrystals.

Background: When using semiconductor quantum dots (QDs) for single-analyte sensing, recognition is commonly achieved through interactions with capping ligands attached to the QDs surface. These ligands form an organic layer that provides stability in solution and assures selectivity by binding the target analyte via surface functional groups. However, a common analytical challenge arises in the subsequent stage of the QD-based sensing scheme. Specifically, in a system combining CdTe and AgInS QDs, the CdTe QD core can effectively dock sodium-2-mercaptoethane sulfonate (MES) ligands, which act as recognition elements for amlodipine. In contrast, the AgInS QD core, may not properly anchor d-penicillamine (PCM) ligands, which bind selectively to olmesartan. This disparity in ligand-QDs affinity may lead to inconsistent excitation-emission fluorescence matrix (EEFM) signals. Such inconsistencies arise when the contribution of the four components present in the system 'CdTe QD-MES-amlodipine & AgInS QD-PCM-olmesartan' deviates from low-rank subspace estimated across different excitation/emission modes.

Results: Upon analyzing a fixed-dosage form sample of amlodipine and olmesartan medoxomil, we demonstrate that the complexity of the mixture signal in EEFM measurements - whether from individual or combined single-analyte QD-based sensing scheme - can be experimentally assessed by determining the low-rank subspace of the analytical signals. Specifically, we evaluated whether the number of fluorescently responsive constituents across different excitation and emission modes matches the four components present in the system 'CdTe QD-MES-amlodipine & AgInS QD-PCM-olmesartan' - a prerequisite for the joint calibration of both analytes. If this condition is not met the analytes must be individually calibrated. Thus, a mixture of single-analyte QD-based sensing scheme, involving different concentrations of amlodipine and olmesartan medoxomil, was used to generate a set of EEFM measurements. A rank-two model was found optimal for emission wavelengths, while a rank-three model was calculated for excitation wavelengths. These results reveal an evident rank-deficient as at least four-components are present in the QD-sensing photoluminescence modulation process.

Significance: The unfolded partial least-squares (U-PLS) model was successfully applied for the determination of amlodipine-net analyte signal (NAS) from binary CdTe QDs fluorescence modulation. In contrast, the olmesartan- NAS from ternary AgInS QDs fluorescence modulation did not allow a combined calibration. Consequently, the analytes must be individually calibrated based on their respective individual mixture signals using either PARAllel FACtor (PARAFAC) analysis or multivariate curve resolution - alternating least-squares (MCR-ALS) model according to the data structure.