This new solution allows for the convenient application of the chemometric-photometric method using the Kalman filter algorithm (Kalman method) to simultaneously det[r]
Trang 1HUE UNIVERSITY HUE UNIVERSITY OF SCIENCES -
NGUYEN THI QUYNH TRANG
RESEARCH ON THE DEVELOPMENT OF
A CHEMOMETRIC METHOD FOR
THE SIMULTANEOUS DETERMINATION
Trang 2HUE UNIVERSITY HUE UNIVERSITY OF SCIENCES
-
NGUYEN THI QUYNH TRANG
RESEARCH ON THE DEVELOPMENT OF
A CHEMOMETRIC METHOD FOR
THE SIMULTANEOUS DETERMINATION
1 Assoc.Prof.Dr TRAN THUC BINH
2 Assoc.Prof.Dr NGO VAN TU
HUE - 2018
Trang 3INTRODUCTION
The term chemometric was first introduced in 1972 by Svante Wold (Swede) and Bruce R Kowalski (American) The establishment of the Chemometric Association in 1974 provided the first definition of chemometrics, the application of mathematical, statistical, graphical methods….for experimental planning, optimize the chemical information extracted from the data set and provide the most useful information from the original data set
Chemometric is widely used in fields such as environmental chemistry, organic chemistry, biochemistry, theoretical chemistry, statistics in chemistry and has especially established an important position in analytical chemistry Analytical chemistry is an effective tool in the fields of science and technology, such as chemistry, biology, agronomy, medicine, food especially in the pharmaceutical industry
Chemometric methods have been used by researchers at domestic and foreign for many years to concurrently analyze a mixture of substances in a variety of subjects, including pharmaceuticals Studies have shown that the most commonly used chemometric methods are the partial least squares (PLS) method, the Polimerase chain reaction method (PCR), classic least squares method (CLS), artificial neural network method (ANN), derivative method, kalman filter method Each method has its own advantages and disadvantages The CLS method can use the entire spectral data to set up the m equations
in n unknowns (m> n).The transformation matrix based on the principle of least squares will produce the results of the error satisfying the requirements However, if there is a lot of noise (or spectral error) in the spectrophotometer and/or when the constituents interact with each other, it produces an optical effect that changes the absorbance of each constituent, this method can not eliminate the noise leading to the analysis results have big errors The ANN method has the disadvantage that the training time is long and it requires a lot of different algorithms, so that when building an analytical model, it requires testing different models to find the optimal network
Trang 4structure Derivative spectrophotometric method does not apply when the sample contains many constituents with absorbing optical spectrum overlapping or similar, since it is difficult to select an appropriate wavelength to determine a particular constituent, or their derivative spectrophotometric still have the same maximum absorption Kalman filter method can eliminate most of the noise and therefore minimize errors, but the disadvantage of this method is that the initial values for the filter must be selected that means must choose the appropriate initial value of the content of analytes in their mixtures and the associated error (expressed by the variance) If the initial values (concentration and variance) do not match, the end result is a large error
In the world there have been some studies applying Kalman filter method to chemometric to simultaneously determine mixtures of 2 or 3 substances in the pharmaceutical However, these studies neither offer a suitable initial value nor cover initial values and are therefore difficult to apply to analytical laboratories In Viet Nam, Mai Xuan Truong has studied the application of Kalman filter method to simultaneously determine the vitamins in pharmaceuticals, rare earth elements However, the author did not introduce how to choose the initial values and thus limited the possibility of applying the proposed method in practice
As a result of these issues, it is clear that the development studies of chemometric-photometric method combined with the use of Kalman filter methods is very necessary, especially to simultaneously determine mixtures of substances difficult to analyze that containing optical absorption spectrophotometer
pharmaceutical samples However, the challenge is to find a suitable solution to select the initial value for the Kalman filter
to produce accurate analysis results (repeatability and accuracy)
or acceptable error At the same time, the analytical process needs to be developed that based on the chemometric-photometric method combined with the Kalman filter so that can be applied conveniently in the field of pharmaceutical
testing in our country For these reasons, the topic "Research on
Trang 5the development of chemometric method for the simultaneous determination of molecular absorption spectrum overlapping and application in drug” was carried out for the following these
purposes:
i) Developed a chemometric-photometric analysis process in combination with Kalman filter method to simultaneously analyze mixtures of 2 and 3 substances with molecular absorption spectra overlapping in pharmaceutical samples;
ii) Apply the process has been built to simultanneously analyze mixtures 2 and 3 substances in some pharmaceuticals are
on the market Vietnam
Master thesis structure
The thesis consists of 184 pages, with 50 tables and 14 figures, of which:
- Table of contents, list of abbreviations, tables and figures:
09 pages
- Introduction: 04 pages
- Chapter 1: Overview 43 pages
- Chapter 2: Content and Research Methods 16 pages
- Chapter 3: Results and discussion 67 pages
- Conclusion 02 pages
- The list of published research results: 01 page
- References: 15 pages, with 127 references
CONTENT THESIS CHAPTER 1 LITERATURE REVIEW
- The Bughe-Lambe-Bia law and Optical properties of optical absorption
+ The Bughe – Lambe - Bia
+ Optical properties of optical absorption
- Some UV-VIS spectrophotometric methods combined with chemometric simultaneously determine the components
with absorption spectrum overlapping each other
+ Vierordt method
+ Derivative spectrophotometric method
+ Full-partial method (least squares method)
+ The least squares method
Trang 6+ Principal Components Regression method
+ Artificial neural network method
+ Kalman Filter Method
- Overview of multi-component pharmaceuticals and research active ingredients
+ Profile of the development of multi-component pharmaceuticals
+ Overview of telmisartan (TEL), hydrochlorothiazide (HYD)
+ Overview of paracetamol (PAR) and caffeine (CAF) + Overview of paracetamol (PAR) and ibuprofen (IB) + Overview of amlodipine besylate (AML), hydrochlorothiazide (HYD), valsartan (VAL)
CHAPTER 2 RESEARCH SUBJECTS AND
METHODOLOGY 2.1 CONTENT
To achieve the objective of the thesis is to contribute to the development of chemometric-photometric method using the Kalman filtering algorithm to apply in pharmaceutical analysis, the research contents include:
1 Study to find the suitable solution to select the initial value (concentration value and initial variance) for the Kalman filter for using the chemometric - photometric method simultaneous determine of molecular absorption spectrum overlapping (mixture contains 2 substances and mixture contains 3 substances)
2 Construct a computer program based on the Kalman filter algorithm on Microsoft-Excel 2016 software with the Visual Basic for Applications programming language, it is possible to quickly calculates of the concentration of photocatalytic absorption spectra overlapped in the study system (containing 2 or 3 substances simultaneously)
3 Verification of the reliability of the analytical method - Chemometric-photometric method using the Kalman filter algorithm (calculated by software program has been built):
Trang 7Comparison of analytical methods with the chemometric- Other photometry (least squares using full spectrum and diffusion method) when analyzing laboratory standard samples (containing 2 or 3 analyzes)
4 Develop a chemometric-photometric analysis using the Kalman filter algorithm (calculated by software program has been built)
5 Apply the analysis process has been built - analysis of multi-component pharmaceutical samples (containing 2 or 3 ingredients) are currently in circulation in Vietnam
2.2 METHODOLOGY
2.2.1 Kalman filter method and calculation program
Based on the theoretical basis, the Kalman filter method and the calculation program are performed according to the following steps (Figure 2.1):
i) Record the spectrum of the analytical solution (laboratory standard solution) and the mixture of analytes, obtaining the spectral data set (optical absorption at selected wavelength k) in the form of a file txt tail (number of wavelengths selected depending on the characteristics of the components in the study);
ii) Enter the mono-particle and compound material data files into a computer software program (programmed in Microsoft-Excel 2016 software) to calculate the (molecular absorption) values of the monomers;
iii) Run the Kalman filter:
- Give the initial initial value, including the first estimate
of the Cest(0) and the covariance of the error Pest(0) (study content (1) will give the initial value);
- Extrapolation of concentration status:
Trang 82.2.2 Minimum squared method using simulan software [2]
Step 1 Prepare standard solutions for each constituent and their mixtures
Step 2: Record the absorption spectra of the standard solution to calculate the absorption coefficient of the constituents: = (ij )mxn
Step 3: Record the optical absorption spectra of the mixed solution, enter the optical absorption matrix measured A = (Ai1)mx1
Step 4: Solve the system of m equations by the least squares method: A = C to find the concentration of C
2.2.3 Derivative spectrophotometric method
Step 1 Prepare standard solutions for each constituent and their mixtures
Step 2: Record the optical absorption spectra and the spectrum, find the appropriate wavelength at which the derivative value of a substance to be analyzed is different from zero or maximum, and the other derivative value is equal to 0 Step 3: After determining the measured wavelength at a certain derivative, proceed to quantify the substances by the benchmark method or add standard
2.2.4 Computer programming method
Trang 9- Calculations to determine the concentration of substances
by Kalman filter method is quite complex, so need to program
on the computer to calculate fast and convenient for users;
- Select open source software is Microsoft-Excel to not infringe copyright;
- Select the language and the tool Visual basic for application;
2.2.5 Data processing method
Application of Microsoft-Excel 2016 software with Data Analysis tool to process experimental data: Calculation of statistical data (arithmetic mean, standard deviation, RSD); Comparison of two repetitions (or two variances), using F (F-test); Comparing two mean values, using t-test; Compare two methods, using paired-t-test
CHAPTER 3 RESULTS AND DISCUSSION
3.1 CHOOSE THE INITIAL VALUE
3.1.1 Select a random initial value
In this way, selecting a random initial value can select any value for the concentration C est(0) and variance P est(0)
[27], [112]
For a mixture containing 2 or 3 substances (a mixture of laboratory standard reagents), the initial values for each substance were randomly selected at a concentration of (0)
est
C = 0,3 µg/mL and variance Pest(0)= 1
Table 3.1 Results of determination of TEL and HYD
concentration in Kalman method with with random selection of
(µg/mL) 9,00 8,00 7,00 6,00 5,00 4,00 3,00 2,00 1,00
Trang 10C
(µg/mL) 0,30 0,30 0,30 0,30 0,30 0,30 0,30 0,30 0,30 RE(%) -97 -96 -96 -95 -94 -93 -90 -85 -70
Table 3.1 shows that with different concentration ratios, between the concentration of the standard solution and the concentration determined to be relatively high RE% error (in the range of 69.7% - 96.7%) The concentration values determined in all mixtures are equal to the initial concentration (0.3 μg / mL)
Table 3.2 Results of determination of AML, HYD and VAL
concentrations in Kalman method with random selection of
Table 3.2 shows that with different concentration ratios, the concentration of the standard solution and the concentration determined were very high (-5.5% - 98.1%) The lowest RE value (-5.5%) corresponds to the standard concentration of 0.325 (close to the initial concentration x = 0.3) The higher the initial concentration, the greater the RE % value
Thus, with the results of the tests in Table 3.1 and Table 3.2, it can be seen that the initial method of selecting the concentration value and the random variance are incomplete, the calculated results still have a relatively large error
3.1.2 Select the assumed initial value
In this study, we investigated a different set of assumed initial value in comparison with previous studies (for aquation
of 2 or 3 substances)
Trang 11- Option 1: Solution 2 (or 3) equation with 2 (or 3)
(unknowns is the substance concentration) at 2 (or 3) wavelengths near each other (the equation depends on the optical absorption and the concentration of the substance in the mixture with the predicted molecular absorption coefficient, Calculates the magnetic spectrum of a monoclonal/monostable solution), determine the concentrations of the substances in the mixture, and take them as the initial concentration values The initial value of the variance is randomly chosen, for example by
1
- Option 2: Select the initial randomized concentration -
(But having purpose) is 0,3 µg/mL (for each substance in a mixture of 2 or 3 substances) But for variance, the initial value for it is not randomly selected, which is calculated by the Horwitz equation: At a concentration of C = 0.3 μg / mL = 3.10-7, calculate the variance by 0.003 and select this value as the initial value
3.1.2.1 For the system two constituents TEL and HYD
Apply Kalman method for mono-spectral data and a mixture of two substances (in the range of 220 nm - 340 nm) with a choice of assumed initial values (according to option 1 and option 2), the results show in Table 3.3 and 3.4
Table 3.3 The results of determination of TEL and HYD
concentrations in the mixture by the Kalman method with the choice of the assumed initial value – Option 1(*)
Mixture H1 H2 H3 H4 H5 H6 H7 H8 H9 TEL
Co (µg/mL) 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00 9,00
C (µg/mL) 0,99 1,99 2,95 3,88 5,03 6,07 7,18 7,99 9,00
RE (%) -0,9 -0,6 -2 -3 -0,6 1 3 -0,1 0 HYD
Co (µg/mL) 9,00 8,00 7,00 6,00 5,00 4,00 3,00 2,00 1,00
C (µg/mL) 8,91 7,84 6,86 6,02 5,06 3,95 3,01 1,98 1,03
RE (%) -1,1 -2,0 -2,0 0,4 1,3 -1,2 0,3 -0,8 3
Table 3.4 The results of determination of TEL and HYD
concentrations in the mixture by the Kalman method with the choice of the assumed initial value – Option 2(*)
Mixture H1 H2 H3 H4 H5 H6 H7 H8 H9 TEL Co (µg/mL) 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00 9,00
C (µg/mL) 0,30 0,30 0,31 0,31 0,32 0,35 0,38 0,42 0,48
Trang 12RE (%) -70,0 -84,9 -89,8 -92,3 -93,5 -94,2 -94,6 -94,7 -94,6 HYD
Co (µg/mL) 9,00 8,00 7,00 6,00 5,00 4,00 3,00 2,00 1,00
C (µg/mL) 0,30 0,31 0,31 0,31 0,31 0,31 0,31 0,31 0,30
RE (%) -96,6 -96,2 -95,6 -94,8 -93,7 -92,2 -89,7 -84,7 -69,7
The results in Table 3.3 and 3.4 show that:
- According to option 1, the Kalman method gives reliable results on the concentration of substances in the mixture with the error of RE <3% (for both TEL and HYD) However, under this option, the implementation is quite complex and depends
on two wavelengths selected to solve the equation that determines the initial concentration values On the other hand, when applied in practice, due to the influence of the matrix, spectral measurements may have greater errors, this option can
be much bigger
- According to option 2, the Kalman method yields large error results, although the initial covariance value is assumed to be more appropriate than the choice of the random variance (1) as in the case before (Section 3.1.1)
- The above results allow us to comment that between concentration and variance, the initial value of the concentration plays a more important (or more decisive) role than the error of the final result (when determined in the Kalman) Obviously, there should be a more appropriate way to choose concentration values
3.1.2.2 For the 3-constituent system AML, HYD ANG VAL
Table 3.5 The result of determination of AML, HYD and VAL
concentrations in the mixture by the Kalman method with a
choice of assumed initial value– Option 1(*)
Co (µg/mL) 0,325 0,65 1,30 5,00
C (µg/mL) 2,794 0,495 1,610 5,910
RE (%) -14,0 -23,8 23,85 18,2 VAL
Co (µg/mL) 4,00 8,00 16,00 5,00
C (µg/mL) 4,796 11,053 29,067 3,949
RE (%) 19,9 38,2 81,7 -21,03
Trang 13(*) C o : Concentration in standard mixed solution; C: Determined concentration
Table 3.6 The result of determination of AML, HYD and VAL
concentrations in the mixture by the Kalman method with a
choice of assumed initial value– Option 2(*)
Mixture H1 H2 H3 H4 AML
Co (µg/mL) 0,250 0,50 1,00 5,00
C (µg/mL) 0,300 0,300 0,282 0,477
RE (%) 20,0 -40,0 -71,8 -90,5 HYD
Co (µg/mL) 0,325 0,65 1,30 5,00
C (µg/mL) 0,301 0,304 0,368 0,443
RE (%) -7,4 -53,2 -71,7 -91,1 VAL
Co (µg/mL) 4,00 8,00 16,00 5,00
C (µg/mL) 0,319 0,454 0,542 0,289
RE (%) -92,0 -94,3 -96,6 -94,2
The results in Tables 3.5 and 3.6 show that:
- According to option 1, except for AML in H2 and H4 mixtures (RE error of 4.5%), the remaining cases had a large error with RE of about 14% - 82% Thus, different from the system two constituents (Their concentration just have error with RE<3%), for the system 3 constituents error is much larger Obviously, as the number of constituents in the system increases, their interaction will be greater, this leads to solve system of 3 equations with 3 unknowns (concentration of substance in the system) will make bigger error Obviously, option 1 only applies to the system 2 constituents On the other hand, this option is also quite complex, since the error of the method depends on the wavelength chosen to establish and solve the equation
- According to option 2, it is similar to the case of the system two constituents, although the introduction of the initial value for the variance is more realistic (as estimated from the Horwitz equation) The error is very large with RE about 7% - 97%)
At this point, we can see that both ways of selecting the initial value for concentration and variance - select a random initial value and set a assumed initial value - have not produced good results (or a large error), unless the initial value of the