Multi-Instrument Calibration with Genetic Regression in UV-Visible Spectroscopy

Abstract

The applicability of genetic regression (GR) to multi-instrument calibration was demonstrated by using several UV-visible spectrophotometers. GR is a calibration technique that optimizes linear regression using a genetic algorithm (GA). Sample spectra of ternary and quaternary mixtures of the pharmaceuticals furaltadone (Fd), doxycycline (Dx), sulfadiazine (Sd), and trimethoprim (Tm) were collected on four different UV-visible spectrophotometers, including one single-beam diode array and three double-beam dispersive instruments. Hybrid calibration models (HCMs) were generated by combining the data collected on multiple instruments into one calibration model as if they had all been collected on a single instrument. For comparison, single-instrument calibration models were also generated for each instrument. Both HCMs and single-instrument models were tested by using a validation set measured on all four instruments. Results obtained from single-instrument models were comparable with a previous study in which partial least-squares (PLS) regression was used for multivariate calibration of these compounds. HCMs for double-instrument cases performed equally well as single-instrument models and slightly worse for the four-instruments models.

Transfer of Multivariate Calibration Models

“…Generally, this requires some sort of systematic variable selection procedure, with the optimization metric specified so that little implicit calibration modeling is embedded in the calibration model. [35][36][37] Wavelength selection can be applied when the spectral differences involve only shifts in the wavelength axis. In this algorithm, spectral variables that exhibit little change in regression coefficients when wavelength is varied are selected for model transfer between instruments.…”

Genetic algorithms in chemometrics and chemistry: a review

“…The same approach is followed by Ozdemir et al [80] for the calibration of multiple instruments, showing that genetic regression is able to produce acceptable models using fewer spectra from the secondary instrument than PLS. The same group also used genetic regression to correct for wavelength drift in single-and multi-instrument calibration [81] and to obtain a multi-instrument calibration in UV-vis spectroscopy [82].…”

Determination of benazepril HCl and hydrochlorothiazide in pharmaceutical preparations using UV-visible spectrophotometry and genetic multivariate calibration method

“…Genetic regression (GR) is a calibration technique that optimizes linear regression models using a genetic algorithm (GA) and has been applied to a number of multi-instrument calibration and wavelength selection problems (14)(15)(16)(17) . GAs are non-local search and optimization methods that are based upon the principles of natural selection (18,19) .…”