Simultaneous chemiluminescence determination of thebaine and noscapine using support vector machine regression

https://doi.org/10.1016/j.saa.2009.12.021Get rights and content

Abstract

In this work, a batch chemiluminescence (CL) method has been proposed for the simultaneous determination of two structurally similar alkaloids, noscapine and thebaine. The method is based on the kinetic distinction of the CL reactions of noscapine and thebaine with Ru(bipy)32+ and Ce(IV) system in a sulfuric acid medium. The least squared support vector machine (LS-SVM) regression was applied for relating the concentrations of both compounds to their CL profiles. The parameters of the model consisting of σ2 and γ were optimized by constructing LS-SVM models with all possible combinations of these two parameters to select the model with the minimum root mean squared error of cross validation (RMSECV) as the best. The parameters of this model were then selected as optimized values. Under the optimized experimental conditions for both compounds, the detection limits obtained using the LS-SVM regression were 0.08 and 0.1 μmol L−1 for noscapine and thebaine, respectively. The proposed method was utilized for the simultaneous determination of the compounds in pharmaceutical formulations and plasma samples with satisfactory results.

Introduction

Opiates were originally available from the opium poppy (Papaver somniferum), native in Asia Minor. Noscapine and thebaine are alkaloids commercially extracted from ‘opium poppies’. Opiates and their derivatives are very potent analgesics [1]. Noscapine is the second most alkaloid in opium, present in concentrations of 2–8% [2].

Several analytical technique have been developed for the simultaneous determination of noscapine and thebaine such as liquid chromatography [3], [4], capillary electrophoresis [5], [6], gas chromatography–mass spectrometry [7], [8], circular dichroism [9], and automated multiple developments TLC [10]. Chemiluminescence analysis (CL) has been attracted because of its sensitivity, relative ease of application, and inexpensive instruments required. However, the method suffers from the lack of selectivity in the case of structurally similar compounds and/or similar reactions. In recent years, a combination of powerful multivariate calibration methods and some analytical techniques has been used to allow for the simultaneous determination of multi-component compounds without previous separation [11].

This work is based on the difference in the chemiluminescence reactions between Ru(bipy)32+ and acidic Ce(IV) in the presence of noscapine and thebaine. Least squares support vector machine (LS-SVM) regression models were constructed to relate CL profiles of the compounds to their concentrations. These models were able to predict noscapine and thebaine concentrations in pharmaceutical formulations and spiked plasma samples.

Support vector machines (SVMs) based on statistical learning theory, introduced by Cortes and Vapnic [12]. Least square support vector machines are an alternate formulation of SVM described by Suykens et al. [13]. Linear least square models attempt to correlate the spectrum xi and reference value yi, of all samples in X. The X represents a [n × p] matrix containing p spectral responses of n samples. The yi and xi denotes the ith column of Y and X, respectively. Y is computed by following equation:Y=XB+b0where B is a [p × 1] vector of coefficients and b0 is the model offset. SVMs regression is based on kernel substitution where X is replaced by a [n × n] kernel matrix (K). K is defined as:κ=k1,1k1,nkn,1kn,nwhere ki,j is defined by the kernel function:ki,j=ϕ(xi)·ϕ(xj)

Kernel functions are defined as any function in input space that corresponds to an inner product in some feature space. The idea of the kernel function is to map data point from input space to high dimensional feature space where points are linearly separable. There are number of kernels that can be used in support vector machines. These include linear, polynomial, radial basis function (RBF) and sigmoid function. The RBF is by far the most popular choice of the kernel types because of their localized and finite responses across the entire range of the real x-axis. Implementation of LS-SVM requires the specification of only two parameters, kernel parameter and gamma, which controls the tradeoff between maximization of margin width and minimizing the number of missclassed sample in the training set. If parameters were not properly selected overfitting and/or under filling phenomena might occur.

Section snippets

Reagents and solutions

All chemical solutions were prepared using reagent grade chemicals and doubly distilled water.

Noscapine and thebaine standard solutions (0.10 mg mL−1) were prepared daily by dissolving 10.0 mg of each compound (Sigma–Aldrich, Germany) in 100 mL of water, stored in a refrigerator and protected from light. The working solutions were prepared by appropriately diluting the stock solution with water.

Ru(bipy)32+ stock solution (1.0 × 10−2 mol L−1) was prepared by dissolving 0.3740 g of

Kinetic profile of CL reaction of Ru(bipy)32+–acidic Ce(IV)-noscapine and thebaine

In the batch CL mode, we found that light emission-time profile for noscapine was narrow with a high CL intensity at about 0.6 s, whereas the CL intensity of thebaine was broad with the highest intensity at about 2.0 s after mixing of the reactants. Kinetic profiles of thebaine and noscapine in the batch mode are shown in Fig. 2. Generally, batch and flow methods are two basic sample-injection modes commonly used for CL analysis. The batch mode has such advantages as minimized reagent consumption

Univariate calibration graph

Under the optimized conditions including Ce(IV) = 3.0 × 10−3 mol L−1 and Ru(II) = 3.0 × 10−3 mol L−1 and 0.085 mol L−1 H2SO4, the CL signal was found to be linear in the concentration ranges of 0.5–5.0 μmol L−1 noscapine and 1.0–50 μmol L−1 thebaine with the linear equations of ICL = 0.585Cnoscapine + 0.515 (r2 = 0.98) and ICL = 0.0194Cthebaine + 0.650 (r2 = 0.9750), respectively, where ICL is CL intensity and C is the concentration of the drugs in μmol L−1. Under the optimized experimental conditions for both compounds, the

Multivariate calibration graph

The LS-SVM regression was constructed using a radial basis function (RBF) as a kernel function. The parameters of the model consisting of γ and δ2 form the critical steps for the performance of the model. Each of the parameters was changed from 1 to 2000 and LS-SVM models were constructed with all the possible combinations of the parameters. A training set of 25 samples was prepared using a full factorial design. Then, for each combination of γ and δ2, the root mean squared error of cross

Interference study

In order to find the selectivity of the proposed method, the interference effects of some organic and inorganic compounds commonly present in human serum and also some redox active compounds were studied by recovering 2.0 μmol L−1 of noscapine and/or thebaine in the presence of potential interfering substances. The tolerance of each substance was taken as the largest amount yielding an error of less than 3σ in the analytical signal of 2.0 μmol L−1 of noscapine and/or thebaine (σ is the standard

Application

To evaluate the applicability of the proposed method, the recovery of noscapine and/or thebaine was determined in plasma as spike sample and in some drug samples. The standard addition method was used for the analysis of the prepared samples. The data are given in Table 4. In addition, the prediction results using LS-SVM were compared with HPLC-UV as an official method [21]. Comparisons of the prediction capabilty of the proposed method and the HPLC method for determination of noscapine and/or

Conclusion

A new CL method was introduced for simultaneous determination of thebaine and noscapine at trace levels without any separation. The chemiluminescence method combined with LS-SVM as a multivariate calibration method was found suitable for the simultaneous determination of the two structurally similar alkaloid compounds. The method is not only a simple procedure, but is also a fast technique. In addition, the method is capable of analyzing both compounds in plasma samples.

Acknowledgements

The authors express their appreciation to the Isfahan University of Technology Research Council and Center of Excellence in Sensor and Green Chemistry for their financial support.

References (21)

  • B.J. Hindson et al.

    J. Pharm. Biomed. Anal.

    (2007)
  • J. Pothier et al.

    J. Chromatogr. A

    (2005)
  • F.A. Aly et al.

    Talanta

    (2001)
  • B. Rezaei et al.

    Spectrochim. Acta A

    (2007)
  • J.A. Murillo Pulgarin et al.

    Anal. Chim. Acta

    (2007)
  • R. Boque et al.

    Anal. Chim. Acta

    (2002)
  • J.M.P.J. Garrido et al.

    Anal. Lett.

    (2004)
  • R. Denooz et al.

    Rev. Med. Liege

    (2005)
  • K. Aramaki et al.

    Anal. Chem.

    (1980)
  • L. Krenn et al.

    Chromatographia

    (1998)
There are more references available in the full text version of this article.

Cited by (19)

  • Classification and determination of sulfur content in crude oil samples by infrared spectrometry

    2022, Infrared Physics and Technology
    Citation Excerpt :

    In this study, a new and simple approach for parallel quantification and qualification of the sulfur content present in crude oils using ATR-FTIR spectroscopy associated with chemometric methods is shown. This new and simple approach will help to estimate the amount of sulfur content in crude oils and classify crude oils based on sulfur content into sweet and sour crude oil, thus to allow better optimization of desulfurization conditions during fuel production [27–42]. The chemometric approaches used for the determination of sulfur content in crude oils are partial least squares regression (PLS-R) and support vector machine regression (SVM-R).

  • Graphitic carbon nitride-graphene nanoplates; Application in the sensitive electrochemical detection of noscapine

    2020, Synthetic Metals
    Citation Excerpt :

    For such studies and to discover the biological function of NOS, it is required to use the simple, sensitive and rapid methods for NOS determination. The determination of NOS in biological fluids is the subject of a few studies [20–27]. A variety of analytical techniques such as chemiluminescence detection [21,22], chromatographic methods including HPLC and GC coupled with sophisticated detection systems [23–27,] have been applied for the monitoring of NOS in a variety of samples.

  • ATR-FTIR spectroscopy and chemometric techniques for determination of polymer solution viscosity in the presence of SiO<inf>2</inf> nanoparticle and salinity

    2019, Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy
    Citation Excerpt :

    According to our review, there is no report on the application of FT-IR and chemometric to measure the viscosity of polymeric solutions in CEOR. A rapid and low cost method is developed for determination of viscosity of polymer solution in the presence of salt and SiO2 nanoparticle based on SVM-R as a non-linear model [28–30] and PLS-R as a linear model [31,32]. Principal component analysis (PCA) is one of several multivariate methods that creates new reduced dimensions of the data and provides precise mathematical estimations of changes along the object and variable vectors without a priori knowledge.

  • A novel highly sensitive thebaine sensor based on MWCNT and dandelion-like Co<inf>3</inf>O<inf>4</inf> nanoflowers fabricated via solvothermal synthesis

    2019, Microchemical Journal
    Citation Excerpt :

    Based on the results, applying the adjusted DNF-Co3O4/MWCNTs/GCE increases the sensitivity whilst decreasing thebaine fouling effects. Table 2 presents a comparison of DNF-Co3O4/MWCNTs/GCE analytical performance created in this research with other methods involved in thebaine analysis [51–55]. The proposed method was inferior in terms of detection limit in comparison to previously reported methods in the literature.

View all citing articles on Scopus
View full text