Premium
A novel computational strategy for the p K a estimation of drugs by non‐linear regression of multiwavelength spectrophotometric pH‐titration data exhibiting small spectral changes
Author(s) -
Meloun Milan,
Bordovská Sylva,
Syrový Tomáš
Publication year - 2007
Publication title -
journal of physical organic chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.325
H-Index - 66
eISSN - 1099-1395
pISSN - 0894-3230
DOI - 10.1002/poc.1235
Subject(s) - chemistry , collinearity , absorbance , linear regression , spectral line , standard deviation , principal component regression , analytical chemistry (journal) , protonation , titration , statistics , mathematics , chromatography , ion , physics , organic chemistry , astronomy
A new computational procedure for the protonation model building of a multiwavelength and multivariate spectra treatment is proposed for the special case of small changes in spectra. The absorbance change Δ i for the i th spectrum divided with the instrumental standard deviation s inst ( A ) represents the signal‐to‐error ratio SER of the spectra studied. The determination of the number of chemical components in a mixture is the first important step for further quantitative analysis in all forms of spectral data treatment. Most index‐based methods of the factor analysis can always predict the correct number of components, and even the presence of a minor one, when the SER is higher than 10. The Wernimont–Kankare procedure in the program INDICES performs reliable determinations of the instrumental standard deviation of the spectrophotometer used s inst ( A ), correctly predicts the number of light‐absorbing components present, and also solves ill‐defined problems with severe collinearity in spectra or very small changes in spectra. The mixed dissociation constants of three drugs, haemanthamine, lisuride, and losartan, including diprotic molecules at ionic strengths of I = 0.5 and 0.01 and at 25°C were determined using two different multiwavelength and multivariate treatments of the spectral data, SPECFIT32 and SQUAD(84) non‐linear regression analyses and INDICES factor analysis, even in the case of small absorbance changes in spectra. The dissociation constant p K a was estimated by non‐linear regression of {p K a , I } data at 25°C: for haemanthamine p K a = 7.28(1) at I = 0.50, for lisuride p K a = 7.86(1) and for losartan p K a,1 = 3.60(1), p K a ,2 = 4.73(1) at I = 0.01. Goodness‐of‐fit tests for the various regression diagnostics enabled the reliability of the parameter estimates found to be proven. PALLAS and MARVIN predict p K a being based on the structural formulae of the drug compounds in agreement with the experimental value. Copyright © 2007 John Wiley & Sons, Ltd.