Sequential acquisition of fluorescence signals with changing fluorophore concentrations. Multivariate Curve Resolution with time measurements assistance.

GABRIEL G. SIANO; SOFIA MORA; AGUSTINA V. SCHENONE; LEONARDO L. GIOVANINI

Santa Fe

Congreso; XIX Chemometrics in Analytical Chemistry; 2024

The sequential acquisition of fluorescence signals in classical Excitation-Emission Matrices (EEMs) with subsequent modeling using algorithms that rely on the multilinear nature of the data, requires that the concentrations of all fluorophores remain relatively constant over time during the acquisition of each EEM. If the concentrations of the fluorophores vary too quickly, as is the case in certain chromatographic runs and in the kinetics of some reactions, these EEMs will deviate from bilinearity, even preventing the use of the potential trilinearity of the third-order data thus generated. This deviation can be severe or mild, depending on the rate of concentration variations of each fluorophore relative to the speed of the sequential spectral scans. However, although any scan can be slow, the reading of each individual point is almost instantaneous. There is no reason to suggest that the signal from these individual points does not preserve its bilinear dependence. Therefore, only within certain time intervals between those required to acquire an entire EEM and an individual point (for example, acquiring half an emission spectrum) can constant concentrations be practically assumed. Instead of implicitly assuming that all data in each EEM were obtained simultaneously, temporally localizing the partial information from each EEM allows for the generation of pseudoEEMs [1], which improve the models with cubes. Although pseudoEEMs allow for the correct modeling of the mentioned third-order data (chromatography, kinetics, etc.), in the case of Multivariate Curve Resolution (MCR), some type of unfolding is additionally required to obtain a second-order array. Being originally third-order data, there are several possible combinations of unfolding, some of which allow the implementation of useful constraints during modeling.Two sets of third-order data were modeled using an implementation of MCR that incorporates timemeasurements from fluorescence records, following a strategy similar to that already reported forchromatographic data [1] with Parallel Factor Analysis (PARAFAC). The second dataset comes fromDiclofenac reaction kinetics. The results obtained suggest that the proposed strategy enables MCR to resolve this type of data. Beyond the inherent rotational ambiguity of bilinear decompositions, in some cases, the solutions found were similar to the unique [2] solutions provided when processing the same data as cubes, without unfolding. However, to obtain appropriate final models, MCR appears to be more dependent on the initial conditions. High degrees of similarity were achieved between resolved and reference spectral profiles.The concentration change profiles, both chromatographic and kinetic, were accurate and with physical meaning. Since a similar time-based strategy was implemented, predictions from calibration models based on MCR results were comparable to those obtained when deriving higher order models from the same data.References[1] Siano GG, Vera Candioti L and Giovanni LL, Chemometrics and Intelligent Laboratory Systems, 2020, 199, 103961[2] Sidiropoulos ND, Bro R, Journal of Chemometrics, 2000, 14, 229-239