This paper presents a preliminary study in building discriminant models from solid-state NMR spectrometry data to detect the presence of acetaminophen in over-the-counter pharmaceutical formulations. The dataset, containing 11 spectra of pure substances and 21 spectra of various formulations, was processed by partial least squares discriminant analysis (PLS-DA). The model found coped with the discrimination, and its quality parameters were acceptable. It was found that standard normal variate preprocessing had almost no influence on unsupervised investigation of the dataset. The influence of variable selection with the uninformative variable elimination by PLS method was studied, reducing the dataset from 7601 variables to around 300 informative variables, but not improving the model performance. The results showed the possibility to construct well-working PLS-DA models from such small datasets without a full experimental design.

A recent study showed that spectral correlation cannot be considered as a reliable method for spot purity of thin-layer chromatography (TLC). The current paper discusses limitations of possible chemometric techniques performed on diode-array detection or multiwavelength densitograms of spots. The primary difference between high-performance liquid chromatography and TLC is the nonlinearity of the detector response. The discussion is based on two examples: a spot of decomposed aspirin and sixteen spots with different concentrations and contaminations of ciprofibrate and clofibric acid. It is concluded that only visual inspection of principal component analysis (PCA) scores can detect inhomogeneity in the cases of nonlinearity, spectral similarity and high overlap. The objective values and other methods can be affected by these phenomena and detect nonlinear artifacts as other compounds, or can be insensitive to a second compound with similar spectra. Therefore, PCA analysis of the spot is the recommended procedure for spot purity checking at the stage of method development.

Summary.The current guidelines of quantitative method development establish a requirement to validate a set of method parameters and the selectivity is one of the most important parameters to be proven. Currently, one of the most frequently used approaches is to measure the spectrum of a spot together with some reference spots and to correlate them. Additionally, several places of the spot (start, middle, end) can be measured and also correlated. In the current paper we perform a simulation done on 170 real spectra with different levels of contamination and noise, establishing the distribution of correlation coefficient. It can be concluded that contamination up to 10% almost cannot be detected by simple correlation measurement, regardless of the noise level. Moreover, at a high noise level there is a very low margin between correlation of different compounds and the correlation of the same (but noisy) spectra. These observations suggest that the spectral correlation cannot be a reliable measure of method specificity and some another chemometric methods need to be searched in further research.

No abstract