Hindawi
International Journal of Analytical Chemistry
Volume 2021, Article ID 5592217, 9 pages
https://doi.org/10.1155/2021/5592217
Research Article
Ultraviolet-Visible Spectroscopy and Chemometric Strategy
Enable the Classification and Detection of Expired Antimalarial
Herbal Medicinal Product in Ghana
Jacob N. Mensah,1 Abena A. Brobbey,1 John N. Addotey ,1 Isaac Ayensu,1
Samuel Asare-Nkansah,1 Kwabena F. M. Opuni ,2 and Lawrence A. Adutwum 2
1Department of Pharmaceutical Chemistry, Faculty of Pharmacy and Pharmaceutical Sciences, College of Health Sciences,
KNUST, Kumasi, Ghana
2Department of Pharmaceutical Chemistry, School of Pharmacy, College of Health Sciences, University of Ghana, Accra, Ghana
Correspondence should be addressed to Lawrence A. Adutwum; ladutwum@ug.edu.gh
Received 2 February 2021; Accepted 18 June 2021; Published 25 June 2021
Academic Editor: Spas D. Kolev
Copyright © 2021 Jacob N. Mensah et al. *is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
To meet the growing demand for complementary and alternative treatment for malaria, manufacturers produce several antimalarial
herbal medicinal products. Herbal medicinal products regulation is difficult due to their complex chemical nature, requiring cum-
bersome, expensive, and time-consuming methods of analysis. *e aim of this study was to develop a simple spectroscopic method
together with a chemometric model for the classification and the identification of expired liquid antimalarial herbal medicinal products.
Principal component analysis model was successfully used to distinguish between different herbal medicinal products and identify
expired products. Principal component analysis showed a clear class separation between all five herbal medicinal products (HMP)
studied, with explained variance for first and second principal components as 37.51% and 26.38%, respectively, while the third principal
component had 18.74%. Support vectormachine classification gave specificity and accuracy of 1.00 (100%) for training set data for all the
products. *e validation set HMP1, HMP2, and HMP3 had sensitivity, specificity, and accuracy of 1.00. HMP4 and HMP5 had
sensitivity and specificity of 0.90 and 1.00, respectively, and an accuracy of 0.98.*e support vector machine classification and principal
component analysismodels were successfully used to identify expired herbalmedicinal products.*is strategy can be used for rapid field
detection of expired liquid antimalarial herbal medicinal products.
1. Introduction the rate of development of resistance. However, the asso-
ciated side effects and adverse drug reactions have made it
*e worldwide mortality attributed to malaria in 2019 was unattractive for some patients [3], as well as cost of treat-
409,000 out of which 384,000 occurred in Africa, with most ment, treatment failures, and accessibility.
of them being children and pregnant women [1]. *ese In such situations, plant medicines have provided a
mortality cases in Africa represent over 93% of all malaria- viable alternative to orthodox medicines for the treatment of
related deaths worldwide. Of the 87 malaria endemic regions malaria. In Ghana, plant medicines remain a major source of
in the world, 28 African countries and India accounted for antimalaria therapy, as a large number of Ghanaians are
95% of malaria cases reported globally [1]. Over the years, observed to patronize herbal antimalaria remedies. Such
concerns have risen about the effectiveness and safety of preference stems from the lower costs of products, perceived
orthodox antimalarial drugs, coupled with the development better efficacy and reduced side effects, and acceptability
of resistance to drugs used in the treatment of malaria [2]. based on peer recommendation. Treatment outcomes from
Artemisinin combination therapy was introduced to reduce the use of these herbal remedies have been positive in a
2 International Journal of Analytical Chemistry
number of cases because medicinal plants have been sources preparations with high volume of solvent and solvent-
of many bioactive compounds including natural scaffolds for wasting separation methods [12, 13].
antimalarial drugs such as quinine and artemisinin [4, 5]. *e ultraviolet-visible (UV-Vis) spectroscopy as a
*ese plant medicines may circumvent the challenges of simple, cost-effective, and nondestructive technique has
parasite resistance and toxicity by the synergistic activity of found applications in environmental, pharmaceutical, and
the several constituent secondary metabolites [6]. other related fields. For example, the British and United
Plant medicines are normally formulated as herbal me- States Pharmacopoeias employ UV-Vis-based methods for
dicinal products (HMPs) with optimum pH and minimal the assay of some pharmaceutical products, as well as ad-
toxicity due to internal buffering effect [5] and relatively low junct method for the identification of certain active phar-
concentration of constituent phytochemicals, respectively. maceutical ingredients.*e technique has also been reported
Currently, the prevalence of use of HMPs in Ghana is about to be useful for the analysis of liquid HMPs [14], and the
76% [7]. Considering the debilitating nature of malaria and the inherent advantages are that the methods are easy to use,
potential for development of organ and neurological compli- enabling laboratories to adopt and effectively implement the
cations in severe cases, there is the need for surveillance and protocols without any extensive training of technical staff. In
continuous quality monitoring of HMPs to safeguard the in- UV-Vis spectroscopy, it has been observed that spectra
tegrity of the products in order to safeguard the general public. obtained from complex mixtures such as liquid HMPs are
However, the efficient quality monitoring of HMPs still re- usually highly convoluted. *is is because the absorption
mains a major challenge. In the case of orthodox medicines, an spectra of several components within the complex samples
assay can be performed to determine the levels of active in- are superimposed on each other. *e authors, however,
gredients and other impurities that may be present. On the envisaged that the application of chemometric techniques to
other hand, HMPs usually have several phytochemical con- the complex UV-Vis data can lead to useful analyses with
stituents that nearly make it impossible to identify all the relevant conclusions on the quality of HMPs.
bioactive compounds and accordingly have them quantified, Chemometric methods involve the use of mathematical
more especially when the products are polyherbal. In spite of and statistical tools to extract useful information from
this, variations in the levels of phytoconstituents in plant complex data [15]. *ese methods have been around for a
materials collected from different sources and at different times while but have become popular lately due to the availability
are a major concern for the monitoring of HMPs for content of easy-to-use software and statistical packages [16, 17].
uniformity, especially in situations where the products are used Chemometric methods are generally used for machine
for the treatment of infectious diseases such as malaria. It is learning and optimization of experiments (i.e., design of
already known that the levels of plant secondary metabolites experiment) [18, 19]. *is work will focus on the former as
are influenced by factors such as growth conditions, time and the latter is beyond the scope of this paper. Machine learning
method of harvesting, and storage condition, as well as the methods are usually termed as supervised or unsupervised
geographical location [8–11]. Since at themoment a lot of these [20]. Unsupervised learning methods such as principal
factors influence the levels of phytoconstituents in medicinal component analysis (PCA) are arguably the most used
plants, and themanufacturing practices are not standardized, it chemometric method [20, 21]. PCA is a dimensionality
becomes important in the interest of the HMP patrons to reduction technique which reveals inherently hidden pat-
develop simple, cost-effective, and efficient system to assure the terns in data. In a PCA model, samples of the same class,
quality of HMPs. which have similar attributes, are clustered closer to each
In view of the difficulty associated with the development other. *us, a confidence ellipse generated around a cluster
of assay methods for HMPs because of the myriad of sec- of samples would show high variability in the score space.
ondary metabolites, the approach by the Food and Drugs Samples deviating from the standard product will be seen
Authority, Ghana, to approve and register HMPs has mainly projecting further away from the center of that cluster. In the
focused on microbial load, toxicity, pH, and some physi- context of our study, this can provide guidance in the de-
cochemical parameters such as ash value and acid value. It is tection of unwholesome or expired HMPs. Supervised
however necessary to have a technique where the protocol learning methods, such as support vector machines (SVM),
involves measurement of some phytochemical components can also be used to classify liquid HMPs in order to dis-
of the HMPs. Some of the challenges associated with such tinguish them from other products. *us, two different
desired methods in resource constrained environments are antimalarial HMPs are expected to be classified into two
the cost of equipment, accessories, and maintenance. In different groups. In addition, variations in the chemical
addition, the methods would be expected to target detection composition due to expiration of the product can also be
and quantification of almost all the phytochemical com- determined as the UV-Vis chemical fingerprint of the
ponents of the polyherbal HMPs or selected markers [12]. product becomes altered. Such alteration, detectible with
*ese methods usually involve techniques such as high- PCA or SVM, can be employed to identify expired HMPs.
performance liquid chromatography (HPLC), gas chroma- *erefore, in this study, we propose a routine for the
tography-mass spectrometry (GC-MS), and capillary elec- classification of liquid antimalarial HMPs using SVM and
trophoresis (CE) which are difficult to find in countries such UV-Vis spectroscopy. We further demonstrate the use of
as Ghana for routine chemical quality monitoring of HMPs. PCA and SVM models to monitor the variations in liquid
Besides the instrumental challenges are tedious sample HMPs and to identify expired products. *is approach has
International Journal of Analytical Chemistry 3
the potential to be expanded to the detection of adulterants was then projected into the model to evaluate the perfor-
in liquid antimalarial HMPs and by extension, other HMPs. mance on an external validation set. *e evaluation was
based on the model’s sensitivity, specificity, and accuracy
2. Experimental [23]. A model’s ability to predict positive samples is true
positive rate/sensitivity (sensitivity� true positives (TP)/
2.1. SamplePreparation. Five liquid antimalarial HMPs were number of positives (NP)). Specificity measures the model’s
obtained from pharmacies and herbal shops in Accra and ability to correctly identify negative samples, also known as
Kumasi, Ghana. *e antimalarials used for this study were true negative rate (specificity� true negatives (TN)/number
packaged in 500mL amber colored plastic bottles. To ensure of negatives (NN)). Accuracy measures the overall true
anonymity, the samples were coded HMP1, HMP2, HMP3, predicting power (accuracy� (TP +TN)/(NP+NN)). *ese
HPM4, and HMP5. *ree different batches each were ob- metrics are scaled from zero to 1, with 0 and 1 being the
tained for each sample. A fourth batch consisting of expired worst and best model, respectively.
HMP4 labelled HMP4X was also obtained from a herbal *e SVM model was further tested on the HP4X, which
shop. are the expired HP4 samples. PCA models were also gen-
A 5mL portion of each sample was pipetted into a 50mL erated and evaluated with the training and validation sets,
volumetric flask. Distilled water was added to the volumetric respectively.
flask to make up to the 50mL mark. *is yielded 5% (v/v)
concentration of the liquid herbal antimalarial in distilled 3. Results and Discussion
water.
*e assessment of quality of HMPs in Ghana has largely
2.2. Data Collection. UV-Vis spectra were collected using a been based on some organoleptic and physicochemical
JENWAY 7315 UV-Vis spectrophotometer (Jenway, UK) parameters including but not limited to color, pH, and
equipped with a Perkin Elmer Spectrum (Spectrum Two, microbial load. However, it is necessary to develop advanced
Version 10.03.09, Serial Number 94133, Waltham, USA). but simple strategies that target the phytoconstituents of
*e samples were analyzed using a 1mL fused silica cuvette HMPs in the evaluation of chemical quality. It is known that
and at a wavelength range of 200 to 700 nm using distilled variation in these secondary metabolites supposed to be
water as blank/reference. At least ten repeated scans were responsible for the biological activities of HMPs exists as a
made for each sample. *e raw data from the spectrum was result of factors including environment, harvesting,
imported into MATLAB R2020b (*eMathWorks®, Natick, manufacturing, storage, and stability. Due to the presence ofMA, USA). PCA and SVMmodels were generated using PLS myriad of chemical compounds in polyherbal products and
Toolbox 8.9 (Eigenvector Research Inc., Manson, WA, the lack of adequate robust analytical methods to check
USA). batch-to-batch consistency in levels of phytoconstituents
and identify expired or decomposed HMPs that may not
show perceptible physical changes, unscrupulous persons
2.3. Data Processing and Analysis. *e data was organized could rebottle and sell substandard and unwholesome
into a matrix of samples in rows and wavelength in columns. products to the general public. *erefore, this study has
*e dataset matrix consisted of 167× 501 (sam- explored the application of UV-Vis spectroscopy and che-
ples×wavelengths). Spectral data was smoothed with a mometric analysis in dealing with the problem. Antimalarial
moving average filter using a window of five. *e spectral HMPs were chosen as a test case due to their high demand
data was subsequently decluttered using generalized least and over-the-counter usage.
squares- (GLS-) based weighting strategy using an alpha Generally, all the UV-Vis spectra obtained from the
value of 0.02 [22]. Each column of the data matrix was mean analyses of the liquid antimalarial HMPs showed maximum
centered while the rows were normalized to 1. absorbance around 230 nm, 280 nm, and 375 nm
*e data was further split into two main groups: 2/3 for (Figure 1(a)), which suggests the presence of compounds
model training and optimization and a third for external with conjugated systems or chromophores and the suit-
validation sets.*e training and validation set data consisted ability of our choice of technique. *is agrees with other
of 111 and 56 samples, respectively. findings that show the presence of chromophoric com-
Discriminant variable (DIVA) test was performed on the pounds in plant products [12].
training set data to identify regions of the spectra providing Due to noise in the dataset, a moving window smoothing
information about the important regions of the data. algorithm was implemented (Figure 1(b)), which showed
*e PCA model was generated with the training and similar spectra characteristics relative to the raw spectra. In
optimizing set data. In the PCA model, the variance order to identify the more informative regions of the spectra,
explained by the model by each component is represented as discriminant variable analysis was performed as previously
a percentage. It is used to demonstrate measure of dis- described [24] to generate a variable selectivity ratio (SR)
crepancy between the model and the data. *us, a higher plot (Figure 2). In this analysis, an SR value less than 10 was
explained variance is desired. *e validation data were considered as one with low discrimination power and, thus,
subsequently projected into the PCA model. was eliminated from the data. *e threshold shows the
*e training data was used to generate SVM classifica- variables which were above the set limit and reduced the
tion models for the 5 classes of samples. *e validation data number of variables from 501 to 324. It must be emphasized
4 International Journal of Analytical Chemistry
4 4
UV-V is spectra of herbal products Smoothened spectra of herbal products
3 3
2 2
1 1
0 0
200 250 300 350 400 450 200 250 300 350 400 450
Wavelength (nm) Wavelength (nm)
(a) (b)
Figure 1: UV-Vis spectra of liquid herbal medicinal products. (a) Raw data. (b) Spectra data smoothed with a moving average filter using a
window of 5.
80
0.04
0.02
60 0
–0.02
40 –0.04
–0.06
20 0.05
PC2 0(26.38 –0.05
0.1
% 0 0.05
0 ) –0.1 –0.05 )
200 250 300 350 400 450 –0.1 1 (37.51%PC
Wavelength (nm) HMP1 train HMP4 train
Threshold HMP2 train HMP5 train
Selectivity ratio HMP3 train
Figure 2: Selectivity ratio plot of UV-Vis spectra of liquid HMP Figure 3: PCA score plot for training set data. A score plot for PC1
obtained from DIVA test showing feature importance to dis- (37.51%) vs. PC2 (26.38%) vs. PC3 (18.74%). HMP1—red circles,
crimination between classes (y-axis) and wavelength (x-axis). HMP2—purple squares, HMP3—green diamonds, HMP4—red
Wavelengths, at which the SR is less than the threshold (red line), pentagrams, and HMP5—blue triangles. Training and validation
were eliminated from the data. data are represented by hollow and filled markers, respectively.
that a fewer number of descriptors with better discrimi- liquid antimalarial HMPs. Subsequently, the external vali-
nating power are much more desirable as they lead to dation set of each HMP class was projected in the model
simpler models [25, 26]. (Figure 4, where HMP1 are red circles, HMP2 are purple
Next, principal component analysis was performed using squares, HMP3 are green diamonds, HMP4 are red pen-
the training set data (only the 324 variables). *e results tagrams, and HMP5 are blue triangles). In Figure 4, samples
show a PCA score plot of PC1 vs. PC2 vs. PC3 (Figure 3).*e used for the training and validation sets are represented by
explained variances for PC1 and PC2 were 37.51% and hollow and filled markers, respectively. Here it is also ap-
26.38%, respectively, while the third principal component parent that those samples that were not used in training the
had 18.74%. *us, a total explained variance of 82.63% was model are also projected into the correct subgroups.
observed. It can be seen in this three-dimensional score *e ability of the PCA model generated to identify
space that the various antimalarial HMPs are all clustered in expired products was also evaluated. Expired products from
separate groups. A 95% confidence ellipse generated around HMP4, labelled as HMP4X, were also projected into the
each cluster shows quite a few of the training set samples model as shown in Figure 4. *ese expired products were
straying out of the cluster. *is demonstrates that UV-Vis clustered into a different score space (represented as black
spectra and PCA models can be used to distinguish various triangles). *is further demonstrates that, using the UV-Vis
Variable selectivity ratio
Absorbance (a.u.)
Absorbance (a.u.)
PC3 (18.74%)
International Journal of Analytical Chemistry 5
0.04
0
–0.04
–0.08
0.05
PC2 0(26.38 –0.05%) –0.1 0 0.05
0.1
–0.1 –0.05
PC1 (37.51%)
HMP1-train HMP4-train
HMP1-valid HMP4-valid
HMP2-train HMP5-train
HMP2-valid HMP5-valid
HMP3-train HMP4X
HMP3-valid
Figure 4: PCA plot for PC1 (37.51%) vs. PC2 (26.38%) vs. PC3 (18.74%) for training set, validation set, and expired products HMP4X.
HMP1—red circles, HMP2—purple squares, HMP3—green diamonds, HMP4—red pentagrams, and HMP5—blue triangles. Black tri-
angles: HMP4X. Training and validation data are represented by hollow and filled markers, respectively.
1.0 1.0
0.75 0.75
0.50 0.50
0.25 0.25
0 0
1 45 90 135 180 1 45 90 135 180
Sample Sample
HMP1-train HMP1-valid HMP1-train HMP1-valid
HMP2-train HMP2-valid HMP2-train HMP2-valid
HMP3-train HMP3-valid HMP3-train HMP3-valid
HMP4-train HMP4-valid HMP4-train HMP4-valid
HMP5-train HMP5-valid HMP5-train HMP5-valid
(a) (b)
Figure 5: Continued.
HP1 pred. prob.
PC3 (18.74%)
HP2 pred. prob.
6 International Journal of Analytical Chemistry
1.0 1.0
0.75 0.75
0.50 0.50
0.25 0.25
0 0
1 45 90 135 180 1 45 90 135 180
Sample Sample
HMP1-train HMP1-valid HMP1-train HMP1-valid
HMP2-train HMP2-valid HMP2-train HMP2-valid
HMP3-train HMP3-valid HMP3-train HMP3-valid
HMP4-train HMP4-valid HMP4-train HMP4-valid
HMP5-train HMP5-valid HMP5-train HMP5-valid
(c) (d)
1.0
0.75
0.50
0.25
0
1 45 90 135 180
Sample
HMP1-train HMP1-valid
HMP2-train HMP2-valid
HMP3-train HMP3-valid
HMP4-train HMP4-valid
HMP5-train HMP5-valid
(e)
Figure 5: SVM class predicted probability for herbal medicinal products HMP1 (a), HMP2 (b), HMP3 (c), HMP4 (d), and HMP5 (e).
HMP1—red circles, HMP2—purple squares, HMP3—green diamonds, HMP4—red pentagrams, and HMP5—blue triangles. Hollow and
filled markers represent training and validation sets, respectively. *e red dashed line represents the discrimination barrier above which
samples are positively predicted as designated by the y-axis label.
spectra and PCA, products that are expired can be easily and it gave consistent results. *e model was generated with
detected. *is is of high importance due to the fact that the training set data and validated using an external vali-
liquid HMPs could go bad without showing perceptible dation set. *e class predicted probability plots for all the 5
variations in their physical appearance and taste. products (HMP1, HMP2, HMP3, HMP4, and HMP5) are
SVM classification models were generated using the 324 shown in Figures 5(a)–5(e), where HMP1 are red circles,
features obtained from the DIVA test. *e model was HMP2 are purple squares, HMP3 are green diamonds,
generated using a radial basis function kernel with cost and HMP4 are red pentagrams, and HMP5 are blue triangles, red
gamma values of 100 and 0.1, respectively. A venetian blind dash lines are discrimination barrier, and training and
cross validation was employed due to structure of the data validation sets are represented by hollow and filled markers,
HP3 pred. prob.
HP5 pred. prob.
HP4 pred. prob.
International Journal of Analytical Chemistry 7
Table 1: Table of results for SVM classification for training and validation sets for all herbal products.
Product ID True positive False negative True negative False positive Sensitivity Specificity Accuracy
Training set
HMP1 24 0 87 0 1.00 1.00 1.00
HMP2 24 0 87 0 1.00 1.00 1.00
HMP3 24 0 87 0 1.00 1.00 1.00
HMP4 20 0 91 0 1.00 1.00 1.00
HMP5 19 0 92 0 1.00 1.00 1.00
Validation set
HMP1 12 0 44 0 1.00 1.00 1.00
HMP2 12 0 44 0 1.00 1.00 1.00
HMP3 12 0 44 0 1.00 1.00 1.00
HMP4 9 0 46 1 0.90 1.00 0.98
HMP5 9 0 46 1 0.90 1.00 0.98
1.0 1.0
0.75 0.75
0.50 0.50
0.25 0.25
0 0
1 32 62 93 124 1 32 62 93 124
Sample Sample
HMP1-train HMP4-train HMP1-train HMP4-train
HMP2-train HMP5-train HMP2-train HMP5-train
HMP3-train HMP4X HMP3-train HMP4X
(a) (b)
1.0 1.0
0.75 0.75
0.50 0.50
0.25 0.25
0 0
1 32 62 93 124 1 32 62 93 124
Sample Sample
HMP1-train HMP4-train HMP1-train HMP4-train
HMP2-train HMP5-train HMP2-train HMP5-train
HMP3-train HMP4X HMP3-train HMP4X
(c) (d)
Figure 6: Continued.
HP3 pred. prob. HP1 pred. prob.
HP4 pred. prob. HP2 pred. prob.
8 International Journal of Analytical Chemistry
1.0
0.75
0.50
0.25
0
1 32 62 93 124
Sample
HMP1-train HMP4-train
HMP2-train HMP5-train
HMP3-train HMP4X
(e)
Figure 6: SVM class predicted probability for herbal medicinal products for training data HMP1 (a), HMP2 (b), HMP3 (c), HMP4 (d), and
HMP5 (e) showing expired samples HMP4X. HMP1—red circles, HMP2—purple squares, HMP3—green diamonds, HMP4—red pen-
tagrams, and HMP5—blue triangles. HMP4X—black triangles. Training and validation data are represented by hollow and filled markers,
respectively. *e red dashed line represents the discrimination barrier above which samples are positively predicted as designated by the y-
axis label.
respectively. In the SVM of Figure 5, class predicted probability developed and applied to the assessment of liquid antima-
closer to zero (black dash lines) indicated less likelihood of larial HMPs.*is method demonstrates the ability of PCA to
samples belonging to the class being predicted. On the other distinguish between different HMPs. In addition, we applied
hand, a class predicted probability close to 1.00 (green dash line) SVM models to UV-VIS spectra of liquid HMPs to classify
indicates that the sample may belong to the class being pre- different antimalarials. Prediction sensitivity, specificity, and
dicted. A class discrimination boundary is represented by a red accuracy of 1.00 (100%) were observed for training set data
dashed line. *e numerical results from Figure 5 are shown in for all the products. With respect to the validation set,
Table 1.*ere were no false negative or false positives leading to sensitivity, specificity, and accuracy of prediction were 1.00
classification sensitivity, specificity, and accuracy of 1.00 in all cases for HMP1, HMP2, and HMP3. However, sen-
(denoting 100%), which demonstrate the power of combination sitivity and accuracy for HMP4 and HMP5 were 0.90 and
of UV-Vis spectrum and SVM for HMPs classification. How- 0.98, respectively. *e SVM method also demonstrated its
ever, in the validation set, two false positiveswere identified from ability to distinguish between wholesome and expired
HMP4 and HMP5. It is normal for a classification model to products.
perform better on a training set data than a validation set.
*e ability of the SVM model to detect expired products Data Availability
was evaluated using the spectra of HMP4X. *e samples in
HMP4X were projected into the model to check, if indeed, the *e data for this project are available at lawrenceadutwum/
model will predict it as HMP4 or others. It can be seen that herbalproducts (github.com).
HMP4X samples are not predicted to be in any of the 5 HMP
groups (Figure 6).*us, the class predicted probability is below Conflicts of Interest
0.5 for all five classes of HMPs. *is is because these products
have undergone some chemical changes even though there is *e authors declare that they have no conflicts of interest.
no perceptible physical change which has altered their UV-VIS
spectra. *is demonstrates that, using UV-Vis spectra and References
SVM classification, expired herbal products can be detected.
*e study scope would be expanded to include batch-to-batch [1] World Health Organization, World Malaria Report 2020,
consistency and stability of HMPs. WHO, Geneva, Switzerland, 2021.
[2] F. Nosten and N. J. White, “Artemisinin-based combination
treatment of falciparum Malaria,” American Society of
4. Conclusion Tropical Medicine and Hygiene, vol. 77, no. 6, pp. 181–192,
2007.
A simple spectroscopic method (UV-Vis spectroscopy) [3] R. Adisa, T. O. Fakeye, and D. Dike, “Evaluation of adverse
together with chemometric models was successfully drug reactions to artemisinin-based combination therapy in a
HP5 pred. prob.
International Journal of Analytical Chemistry 9
Nigeria University community,” Tropical Journal of Phar- [20] M. Alloghani, D. Al-Jumeily, J. Mustafina, A. Hussain, and
maceutical Research, vol. 7, no. 2, pp. 937–944, 2008. A. J. Aljaaf, “A systematic review on supervised and unsu-
[4] H. A. Pawar, “Natural product as a source of lead to the design pervised machine learning algorithms for data science,”
of new drugs,” Natural Products Chemistry and Research, Unsupervised vs Semi-supervised Learning, vol. 47, pp. 3–21,
vol. 2, no. 6, pp. 1–3, 2014, http://paperpile.com/b/rSJ4hz/Tasf 2020.
10.4172/2329-6836.1000156. [21] C. Christophe, “PCA: the basic building block of chemo-
[5] E. G. Azeh, “Malaria and treatment: herbal antimalarials as metrics,” Analytical Chemistry, vol. 29, p. 47, 2012.
alternative to conventional medicine,” Research Journal of [22] Q. Liu, Y. Okui, and Y. Yoshimura, “Generalized least squares
Phytomedicine, vol. 1, p. 1, 2015. model averaging,” Economic Review, vol. 35, no. 8,
[6] S. M. Abdel-Aziz, A. Aeron, and T. A. Kahil, “Health benefits pp. 1692–1752, 2016.
and possible risks of herbal medicine,” Microbes in Food and [23] T. W. Loong, “Understanding sensitivity and specificity with
Health, vol. 36, pp. 97–116, 2016, http://paperpile.com/b/ the right side of the brain,” BMJ, vol. 327, no. 7417,
rSJ4hz/cZL7 10.1007/978-3-319-25277-3_6. pp. 716–719, 2003.
[7] I. A. Kretchy, A. Koduah, K. F. M. Opuni et al., “Prevalence, [24] T. Rajalahti, R. Arneberg, A. C. Kroksveen, M. Berle,
patterns and beliefs about the use of herbal medicinal K. M. Myhr, and O. M. Kvalheim, “Discriminating variable
products in Ghana: a multi-centre community-based cross- test and selectivity ratio plot: quantitative tools for inter-
sectional study,” Tropical Medicine & International Health, pretation and variable (biomarker) selection in complex
vol. 26, no. 4, pp. 410–420, 2021. spectral or chromatographic profiles,” Analytical Chemistry,
[8] D. Donno, R. Boggia, P. Zunin et al., “Phytochemical fin- vol. 81, no. 7, pp. 2581–2590, 2009.
gerprint and chemometrics for natural food preparation [25] A. Amirteimoori, S. Kordrostami, A. Masoumzadeh, and
pattern recognition: an innovative technique in food sup- M. Maghbouli, “Increasing the discrimination power of data
plement quality control,” Journal of Food Science and Tech- envelopment analysis,” International Journal of Operational
nology, vol. 53, no. 2, pp. 1071–1083, 2016. Research, vol. 19, no. 2, p. 198, 2014.
[9] H. C. Andola, K. S. Gaira, R. S. Rawal, M. S. M. Rawat, and [26] L. Angulo-Meza and M. P. E. Linz, “Review of methods for
I. D. Bhatt, “Habitat-dependent variations in berberine increasing discrimination in data envelopment analysis,”
content of berberis asiaticaRoxb. ex. DC. in Kumaon,Western Annals of Operations Research, vol. 116, no. 1-4, pp. 225–242,
Himalaya,” Chemistry & Biodiversity, vol. 7, no. 2, pp. 415– 2002.
420, 2010, http://paperpile.com/b/rSJ4hz/31xH 10.1002/cbdv.
200900041.
[10] I. Zribi, F. Omezzine, and R. Haouala, “Variation in phyto-
chemical constituents and allelopathic potential of nigella
sativa with developmental stages,” South African Journal of
Botany, vol. 94, pp. 255–262, 2014.
[11] A. Lavola, A. Salonen, V. Virjamo, and R. Julkunen-Tiitto,
“Phytochemical variation in the plant-part specific phenols of
wild crowberry (Empetrum hermaphroditum Hagerup)
populations,” Phytochemistry Letters, vol. 21, pp. 11–20, 2017.
[12] D. Kumadoh, K. O. Kwakye, N. Kuntworbe, O. Adi-Dako, and
J. A. Appenahier, “Determination of shelf life of four herbal
medicinal products using high-performance liquid chroma-
tography analyses of aarkers and the systat sigmaplot soft-
ware,” Journal of Applied Pharmaceutical Science, vol. 10,
no. 6, pp. 72–80, 2020.
[13] M. A. Archer, D. Kumadoh, G. N. Yeboah et al., “Formulation
and evaluation of capsules containing extracts of cassia sie-
beriana for improved therapeutic outcome,” Scientific African,
vol. 10, 2020.
[14] K. Kucharska-Ambrożej and J. Karpinska, “*e application of
spectroscopic techniques in combination with chemometrics
for detecting adulteration of some herbs and spices,”
Microchemical Journal, vol. 153, pp. 104–278, 2020.
[15] I. E. Frank and B. R. Kowalski, “Chemometrics,” Analytical
Chemistry, vol. 54, no. 5, pp. 232–243, 1982.
[16] J. Camacho and E. Saccenti, “Chemometrics analysis of big
data,” Comprehensive Chemometrics, vol. 36, pp. 437–458,
2020.
[17] H. Martens, “Quantitative big data: where chemometrics can
contribute,” Journal of Chemometrics, vol. 30, no. 9, p. 559,
2016.
[18] R. G. Brereton, Applied Chemometrics for Scientists, John
Wiley & Sons, Hoboken, NJ, USA, 2007.
[19] M. Otto, Chemometrics: Statistics and Computer Application
in Analytical Chemistry, John Wiley & Sons, Hoboken, NJ,
USA, 2016.