TY - JOUR

T1 - Multivariate calibration transfer between ATR-FTIR and NIR spectrometers: Electronic cigarette refill fluids case study

AU - Юргенсон, Ника

AU - Монахова, Юлия Борисовна

AU - Кирсанов, Дмитрий Олегович

PY - 2025/5

Y1 - 2025/5

N2 - This paper explores an opportunity of transferring multivariate calibration models between the spectroscopic instruments operating on different principles: near infrared spectrometry (NIR) and attenuated total reflection infrared spectrometry (ATR-IR). As an example of such a cross-platform transfer we study the spectroscopic data obtained in the quantification of nicotine, glycerol and propylene glycol in e-cigarette refilling liquids. It is demonstrated that the application of the direct standardization (DS) protocol allowed using the PLS regression models calculated for NIR spectrometer data together with the data obtained with ATR-IR spectrometer and vice versa. The root mean squared error of prediction (RMSEP) values for the transferred data were between 1.0 and 1.5 times higher than that for the original calibrations. The resulted transferred model's accuracy depends a lot on a particular choice of a transfer set and in certain cases the transferred models show comparable accuracy as native models constructed and tested with original instrumental response. This opens up a perspective of developing “universal” cross-platform multivariate calibrations that would be suitable for a particular analytical task regardless the physical operation principles of a measuring device they applied to.

AB - This paper explores an opportunity of transferring multivariate calibration models between the spectroscopic instruments operating on different principles: near infrared spectrometry (NIR) and attenuated total reflection infrared spectrometry (ATR-IR). As an example of such a cross-platform transfer we study the spectroscopic data obtained in the quantification of nicotine, glycerol and propylene glycol in e-cigarette refilling liquids. It is demonstrated that the application of the direct standardization (DS) protocol allowed using the PLS regression models calculated for NIR spectrometer data together with the data obtained with ATR-IR spectrometer and vice versa. The root mean squared error of prediction (RMSEP) values for the transferred data were between 1.0 and 1.5 times higher than that for the original calibrations. The resulted transferred model's accuracy depends a lot on a particular choice of a transfer set and in certain cases the transferred models show comparable accuracy as native models constructed and tested with original instrumental response. This opens up a perspective of developing “universal” cross-platform multivariate calibrations that would be suitable for a particular analytical task regardless the physical operation principles of a measuring device they applied to.

KW - Calibration transfer

KW - Chemometrics

KW - Direct standardization

KW - Electronic cigarettes

KW - Infrared spectroscopy

KW - Near-infrared spectroscopy

KW - Partial least squares regression

UR - https://www.mendeley.com/catalogue/f7666476-790e-300e-9a29-ed19e809991d/

U2 - 10.1016/j.microc.2025.113375

DO - 10.1016/j.microc.2025.113375

M3 - Article

VL - 212

JO - Microchemical Journal

JF - Microchemical Journal

SN - 0026-265X

M1 - 113375

ER -