Abbas et al.

BMC Chemistry (2024) 18:233 BMC Chemistry

https://doi.org/10.1186/s13065-024-01339-4

RESEARCH Open Access

Simultaneously quantifying a novel

five‑component anti‑ migraine formulation
containing ergotamine, propyphenazone,
caffeine, camylofin, and mecloxamine using UV
spectrophotometry and chemometric models
Ahmed Emad F. Abbas1* , Nahla A. Abdelshafi2, Mohammed Gamal3, Michael K. Halim1,
Basmat Amal M. Said4, Ibrahim A. Naguib5, Mohmeed M. A. Mansour6, Samir Morshedy7 and Yomna A. Salem8

Abstract
This study presents a new method for simultaneously quantifying a complex anti-migraine formulation containing
five components (ergotamine, propyphenazone, caffeine, camylofin, and mecloxamine) using UV spectrophotometry
and chemometric models. The formulation presents analytical challenges due to the wide variation in component
concentrations (ERG: PRO: CAF: CAM: MEC ratio of 0.075:20:8:5:4) and highly overlapping UV spectra. To create a com-
prehensive validation dataset, the Kennard-Stone Clustering Algorithm was used to address the limitations of arbitrary
data partitioning in chemometric methods. Three different chemometric models were evaluated: Classical Least
Squares (CLS), Partial Least Squares (PLS), and Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS).
Among these, MCR-ALS demonstrated excellent performance, achieving recovery values of 98–102% for all compo-
nents, accompanied by minimal root mean square errors of calibration (0.072–0.378) and prediction (0.077–0.404).
Moreover, the model exhibited high accuracy, with relative errors ranging from 1.936 to 3.121%, bias-corrected mean
square errors between 0.074 and 0.389, and a good sensitivity (0.2097–1.2898 μg ­mL−1) for all components. The
Elliptical Joint Confidence Region analysis further confirmed the predictive performance of the models, with MCR-ALS
consistently showing the smallest ellipses closest to the ideal point (slope = 1, intercept = 0) for most analytes, indicat-
ing superior accuracy and precision. The approach’s sustainability was rigorously assessed using six advanced met-
rics, validating its environmental friendliness, economic viability, and practical application. This approach effectively
resolves complex pharmaceutical formulations, contributing to sustainable development objectives in quality control
processes.
Keywords UV Spectrophotometry, Chemometric models, Kennard stone clustering algorithm, Multi-component
pharmaceutical analysis, Sustainability assessment

*Correspondence:
Ahmed Emad F. Abbas
dr.ahmedeemad@gmail.com; ahmed.emad.pha@o6u.edu.eg
Full list of author information is available at the end of the article

© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0
International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or
parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To
view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by-​nc-​nd/4.​0/.
Abbas et al. BMC Chemistry (2024) 18:233 Page 2 of 21

Introduction strategic implementation of chemometrics solutions. By

The pursuit of sustainable, cost-effective, and environ- harnessing the power of chemometrics, UV–visible spec-
mentally conscious practices in pharmaceutical analysis troscopy is transformed into a practical green analytical
has become a paramount imperative. The recent market tool, ideally suited for efficient and sustainable pharma-
launch of a novel fixed-dose anti-migraine formulation ceutical quality control workflows [9–12].
of ergotamine (ERG), propyphenazone (PRO), caffeine A common pitfall in numerous recent chemometrics
(CAF), camylofin (CAM), and mecloxamine (MEC) has research studies is the reliance on random partitioning
posed a considerable analytical problem. To date, only to segregate the sample data into training and validation
a single analytical method has been reported for this subsets [13, 14]. Even though random splitting is simple,
complex mixture [1], and it falls short of aligning with it can result in validation sets that do not adequately rep-
the progressive principles of green analytical chemis- resent the full sample space, potentially leading to biased
try (GAC) and white analytical chemistry (WAC). The outcomes or an overestimation of the model’s accuracy.
existing method’s reliance on harmful solvents, intri- This outcome directly contradicts the sustainable objec-
cate multi-stage procedures, expensive equipment, and tives of resource efficiency and reliability. To surmount
extensive laboratory infrastructure not only contradicts this formidable challenge, our research harnesses the
sustainable practices but also poses significant risks to Kennard Stone Clustering Algorithm (KSC), an advanced
human health, the environment, and practical economic statistical approach, to construct truly representative
considerations, necessitating a paradigm shift towards validation sets [15]. This method partitions the modeled
the development of innovative analytical approaches that variable space into distinct clusters, ensuring that vali-
embrace these emerging disciplines [2–5]. dation samples encompass the full spectrum and distri-
The quantitative determination of this fixed-dose com- bution of each variable. This meticulous clustering and
bination presents significant analytical challenges, pri- sample selection strategy comprehensively covers the
marily due to the vastly different concentrations of its sample space, enabling an unbiased assessment of the
components. The challenging ratio of ERG (0.075), PRO chemometric model’s predictive capabilities across all
(20), CAF (8), CAM (5), and MEC (4) requires a method data domains. KSC enhances analytical sustainability by
with high sensitivity and a wide linear range. ERG, pre- minimizing material consumption and waste generation
sent in the lowest concentration, demands exceptional while bolstering reliability through fewer, strategically
sensitivity for accurate quantification. Its instability and distributed validation samples.
light sensitivity further complicate analysis. PRO, the In this context, our study addresses five crucial objec-
highest concentrated component, necessitates careful tives: (1) Developing an environmentally benign, highly
sample preparation to avoid detector saturation while sensitive, and cost-effective chemometric quantification
maintaining the working range. CAF, a common ingre- method for ERG, PRO, CAF, CAM, and MEC without
dient in many formulations, may introduce potential prior separation, marking a significant improvement over
interferences. CAM and MEC, present in intermediate the existing method; (2) Positioning chemometrics as a
concentrations, add to the complexity of the mixture due sustainable analytical alternative to traditional methods,
to potential spectral overlaps, as shown in (Fig. 1). More- facilitating economical and routine quality control pro-
over, the structural similarities and potential interactions cesses; (3) Systematically leveraging KSC to construct
between these compounds can lead to matrix effects, validation sets that rigorously assess predictive reliabil-
further complicating their simultaneous determina- ity across all concentration ranges, minimizing bias and
tion. These factors collectively underscore the need for a overfitting; (4) Conducting a comprehensive greenness
robust, sensitive, and selective analytical method capable evaluation using state-of-the-art tools including the Car-
of accurately quantifying all five components in a single bon Footprint Analysis (CFA), National Environmental
analysis. Method Index (NEMI), Complementary Green Analyti-
In this framework, UV–visible (UV–Vis) spectropho- cal Procedure Index (Complex GAPI), and Analytical
tometry presents itself as an effective approach, aligning Greenness Metric (AGREE) analysis, providing a multi-
with the sustainability goals of modern analytical chem- faceted assessment of environmental impact; and (5) Pio-
istry. It offers a cost-efficient solution by utilizing afforda- neering "whiteness" and "blueness" assessments through
ble reagents and simple apparatus, including cuvettes and the novel application of the Blue Applicability Grade
light sources, while producing minimal hazardous waste. Index (BAGI) and Red–Green–Blue 12 (RGB12) metrics,
This technique adheres to the tenets of both GAC and quantifying the practicality, sustainability, analytical per-
WAC [6, 7]. However, significant spectral overlaps fre- formance, and cost-effectiveness advantages over existing
quently make it difficult to directly quantify pharmaceu- methods. This innovative approach heralds a paradigm
ticals [8], a challenge that can be surmounted through the shift towards eco-friendly, cost-effective, and practical
Abbas et al. BMC Chemistry (2024) 18:233 Page 3 of 21

Fig. 1 The chemical structure of ERG, PRO, CAF, CAM, and MEC

analytical strategies, aligning with sustainable develop- To optimize spectral resolution and data density, we con-
ment principles and quality control excellence in phar- figured the spectrophotometer with a slit width of 1.0 nm
maceutical analysis. and implemented a fine-grained 0.1 nm sampling inter-
val. Measurements were conducted in fast scan mode
Experimental with a single sweep to balance efficiency and data quality.
Analytical instruments and apparatus In addition, a high-precision Shimadzu analytical balance
Gathering spectral information was accomplished utiliz- (AGE-220) was utilized for accurate sample prepara-
ing a Shimadzu UV-1800 double-beam spectrophotome- tion, and an ultrasonic bath for extraction was provided
ter equipped with 1 cm quartz cells. Coupled with quartz by Julabo Labortechnik of Germany. We processed the
cuvettes of 1 cm path length. The instrument’s operation data and performed chemometrics analysis using Matlab
was managed through UV-Probe software (version 2.42). R2013a with PLS Toolbox v2.0 and MCR-ALS Toolbox
Abbas et al. BMC Chemistry (2024) 18:233 Page 4 of 21

(an allowed program accessible at http://​www.​mcrals.​ instrumentation while accurately reflecting the antici-
info). To implement KSC, we used the built-in scripts of pated concentration levels in pharmaceutical formula-
MATLAB R2013a (version 8.2.0.701), which make use of tions, thereby ensuring robust analytical performance
advanced statistical algorithms. across the relevant concentration domain.

Chemicals and materials Methodology and design of experiment

All chemicals were of analytical grade. We got ERG, A methodical experimental outline is essential for attain-
PRO, CAF, CAM and MEC as reference standard com- ing spectral data that is both representative and abundant
pounds from the National Organization for Drug Con- in information. A multiple-level, and multiple-factor cali-
trol and Research of Egypt. They were certified to be bration set consisting of twenty-five mixtures was devel-
99.58%, 99.39%, 99.75%, 99.63% and 99.72% pure, respec- oped, following the strategy presented by Brereton et al.

St.Louis, USA. The ­SPASMOMIGRAINE® tablets (batch

tively. Ethanol was bought from Sigma Aldrich Co., [16]. As a result, a set of 25 mixes was created for cali-
bration with different ERG, PRO, CAF, CAM and MEC
no: S2456) made by Kahira Pharmaceuticals & Chemi- concentrations ranging from 1 to 5, 10–30, 4–12, 1–9,
cal Industries Company (Cairo, Egypt) and containing and 2–6 μg ­mL−1 for ERG, PRO, CAF, CAM and MEC,
0.75 mg ERG, 200 mg PRO, 80 mg CAF, 50 mg CAM respectively. KSC was used to build a validation set that
and 40 mg MEC in each tablet, was bought from a local reliably validates the model and provides an illustrative
pharmacy. sample of the concentration space. To construct a robust
validation set, we implemented a stratified sampling
Standard solutions approach, splitting the concentration range into thirteen
We prepared primary stock solutions for ERG, PRO, CAF, equally probable strata. From each stratum, we selected
CAM, and MEC at a concentration of 1000 μg ­mL−1. This one unique mixture, resulting in a comprehensive set of
process involved precise weighing of 100 mg of each ref- 13 distinct validation samples. This sampling strategy
erence standard using an analytical balance, followed by ensures thorough coverage of the entire concentration
dissolution in ethanol within 100 mL volumetric flasks. domain, enhancing the reliability of our method valida-
The solutions were then brought to volume with meticu- tion. Sample preparation was executed with precision
lous attention to the meniscus. The stock solutions exhib- using calibrated micropipettes for accurate volume trans-
ited stability for one month when refrigerated at 4 °C. fers. Ethanol served as our solvent of choice, balancing
To attain the requisite concentration ranges for analy- effective analyte dissolution with environmental consid-
sis, we performed daily dilutions of the stock solutions erations. All mixes were formulated in 25 mL volumet-
using ethanol to create fresh working standard solutions, ric flasks, ensuring consistent and precise final volumes.
ensuring optimal sample integrity. Spectral acquisition was performed using 1 cm quartz
cuvettes, chosen for their superior optical properties and
Working range and spectral properties chemical inertness. We recorded absorption spectra over
To elucidate the spectral characteristics of each analyte, 200–400 nm wavelength range, capturing the full UV
we acquired individual UV absorption spectra for ERG, spectral fingerprint of our analytes. To account for back-
PRO, CAF, CAM, and MEC across the 200–400 nm ground absorbance and ensure measurement accuracy,
wavelength range. For this spectral profiling, we utilized we employed ethanol as a blank reference. This carefully
standard solutions with concentrations of 1 μg ­ mL−1 designed experimental approach not only meets the rig-
−1 −1 −1
ERG, 20 μg ­mL PRO, 8 μg ­mL CAF, 5 μg ­mL CAM, orous standards of analytical method development but
and 4 μg ­mL−1 MEC. These specific concentrations were also demonstrates our commitment to sustainable and
strategically selected to bridge the gap between phar- responsible scientific practices in pharmaceutical analy-
maceutical formulation ratios and the linear dynamic sis. Lower and higher spectral ranges were thrown out
range of our analytical method. To establish and vali- because they had too much noise or no signals, making
date the method’s working range, we conducted a com- the working spectral data matrix 220–350 nm with a
prehensive analysis of ERG, PRO, CAF, CAM, and MEC resolution of 1 nm (131 data points), as shown in (Fig. 2).
mixtures, scanning from 200 to 400 nm. The concen- This spectrum information was used to create and verify
tration working ranges were carefully optimized for the chemometric models.
each analyte: 1–5 μg ­mL−1 for ERG, 10–30 μg ­mL−1 for
PRO, 4–12 μg ­mL−1 for CAF, 1–9 μg ­mL−1 for CAM, Models’ development and refinement
and 2–6 μg ­mL−1 for MEC. These meticulously chosen Our study employed three distinct chemometric
intervals served a dual purpose: they aligned with the regression approaches: classical least squares (CLS),
linear response capabilities of our spectrophotometric partial least squares (PLS), and multivariate curve
Abbas et al. BMC Chemistry (2024) 18:233 Page 5 of 21

0.500
PRO (20 g/mL)

CAF (8 g/mL)
0.400
CAM (5 g/mL)

ERG (1 g/mL)
0.300

MEC (4 g/mL)

0.200

0.100

0.000
220.00 250.00 300.00 350.00
Wavelength (nm)
Fig. 2 The zero-order absorption spectrum of ERG, PRO, CAF, CAM, and MEC

resolution-alternating least squares (MCR-ALS). To addressing the complex analytical challenges presented
ensure model robustness and mitigate overfitting, we by our multi-component pharmaceutical formulation.
meticulously optimized each calibration model using a
comprehensive 25-mixture design calibration set. For
the CLS model, we implemented a wavelength-specific Analytical performance metrics
regression strategy, eschewing the integration of results The predictive power, robustness, precision, accuracy,
into latent variables (LVs). To enhance model perfor- and sensitivity of the optimized models were assessed by
mance, we utilized a moving window wavelength selec- computing a number of critical parameters [17]. Using
tion technique. We systematically evaluated window the calibration set spectra, root mean square error of
widths from 5 to 30 nm via cross-validation to determine calibration (RMSEC), standard error of calibration (SEC),
the ideal spectral smoothing level that effectively reduces and RMSECV were calculated as pointers of the model’s
noise while preserving crucial quantitative information. fitting and predictive power on the calibration set.
In developing the PLS model, we methodically varied Relative root mean square error of prediction (RRM-
the number of LVs from 1 to 10. The optimal LV count SEP) was used to quantify predictive accuracy for valida-
was determined using Venetian blinds cross-validation, tion set performance. Using the root mean square error
with the root mean squared error of cross-validation of prediction (RMSEP), the generalization ability of the
(RMSECV) serving as our performance metric. This model was evaluated overall. Additionally, bias-corrected
approach allowed us to strike a balance between model mean square error of prediction (BCRMSEP) was used to
fit and complexity. For the MCR-ALS model, constraint assess the precision and predictability of fresh samples.
optimization played a pivotal role in calibration refine- To compute RMSEP, RMSECV, and RMSEC, we use the
ment. We applied non-negative least squares (nnl) or following equations [17]:
non-negativity constraints to both spectral profiles and 
n
concentration. This strategy facilitated the attainment yi)2
i=1 (yi − 
of satisfactory parameters with minimal iterative cycles, RMSE =
n
enhancing computational efficiency. Through this rigor-
ous optimization process, we aimed to develop highly The subsequent equations were employed to calculate
reliable and accurate chemometric models capable of the further metrics of merit:
Abbas et al. BMC Chemistry (2024) 18:233 Page 6 of 21

n
i=1 (yi yi)
− for 15 min. The volume was filled with ethanol, thor-
Bias = oughly mixed, and then diluted with ethanol to obtain
n
concentrations of 7.5 μg ­mL−1 ERG, 2000 μg ­mL−1 PRO,

n 800 μg ­mL−1 CAF, 500 μg ­mL−1 CAM, and 400 μg ­mL−1
i=1 (yi yi − bias)2
− MEC. Subsequently, the mixture underwent filtration
SEC =
n−1 utilizing a 0.45 μm membrane filter. Appropriate aliquots
of the clear filtrate were diluted with ethanol in volumet-
 ric flasks (10 mL). At this stage, while the concentrations
n
1
n i=1 (yi yi)2
− of PRO, CAF, CAM, and MEC fell within their respec-
RRMSEP% = X100
yi tive linear ranges, the ERG concentration was below the
limit of quantification. To address this, a standard addi-
n tion approach was employed. A 1.5 μg ­mL−1 ERG stand-
i=1 (yi − yi)2 ard solution was added to each determination, bringing
BCRMSEP = − (bias)2
n the total ERG concentration within the working range.
The outcome of the cross-validation (RMSECV), vali- These diluted sample solutions’ absorption spectra were
dation (RMSEP), and calibration (RMSEC) procedures is obtained between 200 and 400 nm in relation to an etha-
denoted by the value yi. The variables n, and yi represent nol blank. The spectra were examined using the proposed
the overall number of samples, and the experimental out- chemometric models to calculate the concentrations of
come for sample i, respectively. ERG, PRO, CAF, CAM, and MEC in the pharmaceutical
To assess the accuracy, (2, 3, and 4 μg ­mL−1) for ERG, mixture. By subtracting the additional standard concen-
(15, 20, and 25 μg ­mL−1) for PRO, (6, 8, and 10 μg ­mL−1) tration from the total measured concentration, the true
for CAF, (3, 5, and 7 μg ­mL−1) for CAM and (3, 4, and ERG concentration in the samples was determined. The
5 μg ­mL−1) for MEC were determined in triplicate at accuracy was measured by the addition of spiked stand-
three different concentrations inside the linear range, and ards in triplicate at 4 distinct levels of concentration. We
the percent recoveries (%R) were calculated. The percent calculated RSD% and R%.
relative standard deviation (%RSD) quantified precision
for repeatability (intra-day) and intermediate (inter-day) Results and discussion
precision. Three different samples were tested three Green solvent selection
times, once on the same day and three times on three dif- Evaluation using GSST tool
ferent days, (2, 3, and 4 μg ­mL−1) for ERG, (15, 20, and An essential component of achieving sustainability in
25 μg ­mL−1) for PRO, (6, 8, and 10 μg ­mL−1) for CAF, (3, analytical procedures is choosing the right green sol-
5, and 7 μg ­mL−1) for CAM and (3, 4, and 5 μg ­mL−1) for vents. A number of pharmaceutical companies have pub-
MEC. Robustness was evaluated through experiments lished solvent sustainability recommendations that use
employing marginally different wavelength intervals data from safety data sheets (SDS) to analyze the advan-
(0.9 nm as opposed to 1 nm), scan speeds (medium as tages and disadvantages of recommended solvents. These
opposed to fast scan), and spectral bandwidths (0.8 nm companies include GlaxoSmithKline, Sanofi, and Pfizer,
as opposed to 1 nm slit width). The method’s robustness among many others. We used a solvent selection tool that
was shown by its capacity to endure these alterations. To was recently introduced by Larsen et al. [18] to help us
evaluate sensitivity, net analyte signals were utilized to with this process. Using this chemometric tool, we can
compute the limits of quantification (LOQ) and detec- compare and evaluate solvents quantitatively according
tion (LOD). This inclusive validation offers a detailed to important eco-friendliness criteria like health, safety,
insight into how well, accurately, precisely, sensitively, environmental impact, and waste management. For every
and robustly the models we made can predict the analysis solvent, it computes a composite greenness score (G),
of pharmaceutical mixtures. where higher scores correspond to more sustainability.
We used this tool with seven different polar solvents,

­PASMOMIGRAINE®
Pharmaceutical dosage forms analysis including acetonitrile, ethyl acetate, ethanol, hexane,
The content of 10 tablets of S chloroform, and water. Water, acetonitrile, ethyl acetate,
tablets was accurately weighed, finely powdered, and ethanol, and methanol had much higher G scores than
thoroughly mixed. A quantity of powder equivalent to harmful solvents like hexane and chloroform, according
0.75 mg ERG, 200 mg PRO, 80 mg CAF, 50 mg CAM, and to the software analysis (Fig. 3). To illustrate, acetonitrile,
40 mg MEC was transferred into a volumetric flask with a water, ethyl acetate, ethanol, and methanol all achieved
capacity of 100 mL. To extract the drugs into the solution, high G scores —7.3, 6.7, 6.6, 5.8, and 5.8, respectively,
about 80 mL of ethanol was added, and it was sonicated according to positive evaluations in the areas of health
Abbas et al. BMC Chemistry (2024) 18:233 Page 7 of 21

Fig. 3 a G scores by GSST attained by the online tool. b Spider diagram attained by SDAGI

effects, safety, environmental effects, and waste disposal. chemistry philosophies by using safer, more environmen-
As illustrated in (Fig. 3). We chose acetonitrile, water, tally friendly chemicals.
ethyl acetate, ethanol, and methanol as more environ-
mentally friendly solvents for additional evaluation based SDAGI assessment
on these greenness evaluations. This fits in with our Even though tools such as GSST facilitate initial sol-
plan to create an analytical technique that follows green vent selection, a more thorough evaluation of reagent
Abbas et al. BMC Chemistry (2024) 18:233 Page 8 of 21

greenness utilizing comprehensive investigational data modeling enables the extraction of latent concentration
is required. When assessing solvent and reagent green- data from the intricate spectral matrix.
ness, the SDAGI tool offers a useful qualitative technique An essential phase in model building entailed the care-
that is dependent on SDS data [19, 20]. This approach ful selection of spectral areas to enhance valuable signals
uses Safety, Health, and Environmental (SHE) data on while reducing the impact of noise and spectral artifacts.
the properties and effects of reagents. Greenness rat- Using an iterative optimization procedure, including
ings based on important criteria can be computed and signal-to-noise ratios, probable interferents, and prelimi-
visual spider diagrams can be made to evaluate a solvent’s nary model performance measurements, we established
greenness. Marks from -5 to + 5, derived from the five an ideal spectral window of 220–350 nm, yielding 131
assessment subcategories of fire safety, general proper- data points per spectrum. The resultant data matrix was
ties, health impact, stability, and odor are displayed in the utilized as input for the development and optimization of
hierarchical spider plot. Secondary spider charts show CLS, PLS, and MCR-ALS models employing MATLAB
subgroup specifics, as (Fig. 3) demonstrates. Using a spi- software.
der diagram to evaluate and compare chemicals visually To validate the fundamental assumption of spectral
is a straightforward way. The percentage of the avail- additivity underpinning CLS and MCR-ALS models, we
able data that was used to generate the Greenness Index conducted a rigorous additivity assessment. This evalu-
outputs is shown in the "Greenness Index Table". Using ation involved the preparation of a synthetic standard
SDAGI, we examined SDS data and discovered that, as mixture mirroring the concentrations found in the phar-
indicated in (Table S1), methanol and acetonitrile had maceutical formulation. We then acquired and compared
scores of -0.12 and -0.30, while ethanol and ethyl acetate three distinct spectral profiles: (1) The UV spectrum of
showed larger safety zones and elevated average green- the synthetic standard mixture, (2) the spectrum of the
ness scores (1.33 and 1.44 for ethanol and ethyl acetate, actual pharmaceutical formulation, and (3) a compos-
respectively). Water, ethyl acetate, and ethanol were ite spectrum generated by the mathematical summa-
therefore chosen as the more environmentally friendly tion of individual component spectra at their respective
solvents for more testing in order to determine which concentrations.
green solvent would be best. However, after evaluating Comparative analysis revealed remarkable concord-
the UV spectra of ERG, PRO, CAF, CAM and MEC in the ance among these spectra, with deviations not exceeding
3 solvents, we found that ethanol provided a higher UV 2% at any wavelength across the analytical range. Spectral
response and better spectral shape compared to water features, including peak positions and relative intensities,
and ethyl acetate. Considering the superior UV analytical demonstrated high consistency. Notably, we observed no
performance of ethanol along with its environmentally emergent peaks or spectral shifts in the mixture spec-
friendly nature, we concluded that ethanol is the most tra relative to the calculated composite, indicating the
suitable green solvent for further studies involving ERG, absence of significant inter-analyte interactions (Fig. S1).
PRO, CAF, CAM and MEC. Statistical evaluation using a paired t-test to compare
absorbance values between the synthetic mixture spec-
trum and the mathematically derived composite showed
Development of CLS, PLS, and MCR‑ALS chemometric no statistically significant differences (p > 0.05) across all
models wavelengths. Furthermore, residual analysis of the dif-
The complex spectral profiles of ERG, PRO, CAF, CAM ference spectrum between these two profiles revealed
and MEC exhibit significant overlap in their UV absorp- only stochastic noise, devoid of systematic deviations.
tion spectra, as evidenced in (Fig. 2). This spectral These findings provide robust support for the additivity
congestion precludes direct quantification through con- assumption, confirming the absence of spectral-altering
ventional univariate methods, necessitating the applica- interactions among the analytes in the mixture state.
tion of advanced chemometric models [21–23]. To tackle This validation substantiates the applicability of CLS and
this analytical challenge, we developed and optimized MCR-ALS models alongside PLS for this complex ana-
three multivariate chemometric models using the com- lytical system.
plete spectrum data of the five-component mixture.
Spectral data acquisition was performed over the range Calibration set design
of 200–400 nm, capturing the characteristic UV–vis To find the best combination of the three components for
fingerprints of the analytes. While these spectral signa- the calibration set, we used an experimental design with
tures encode the requisite concentration information, the multiple levels and factors that Brereton et al. [16] sug-
substantial overlap impedes straightforward correlation gested to create 25 different mixtures. Improved model
between absorbance and analyte levels. Chemometric accuracy is achieved by using this approach to create
Abbas et al. BMC Chemistry (2024) 18:233 Page 9 of 21

analyte concentration profiles that are not correlated. Table 1 The five-level five-factor experimental design of 25
Different concentrations were tested at five levels, with calibrations mixtures together with the 13 validation set mixtures
0 being the middle point and the others being −2, −1, used in the chemometric methods
0, + 1, and + 2 As shown in (Table 1). The analyte pro- Mix no Calibration set (µg/mL)
files’ lack of correlation and the structured variation in
ERG PRO CAF CAM MEC
concentrations help the model more successfully sepa-
rate each component’s spectrum impact. As a result, 1 3 20 8 5 4
the quantification accuracy is improved, and overfitting 2 3 10 4 9 3
is prevented. This saved time and money in the lab by 3 1 10 12 3 6
reducing the quantity of materials used, chemical waste, 4 1 30 6 9 4
and calibration sample requirements, consequently, mak- 5 5 15 12 5 3
ing the method more environmentally compatible. In 6 2 30 8 3 3
general, the calibrations design of multilevel and multi- 7 5 20 6 3 5
factor enhanced reliability and accuracy through careful 8 3 15 6 7 6
modeling and selection of the complete analyte concen- 9 2 15 10 9 5
tration spectrum. This aligns with the fundamental ideas 10 2 25 12 7 4
that guide the creation of ecologically friendly analytical 11 4 30 10 5 6
procedures. 12 5 25 8 9 6
13 4 20 12 9 2
Design and optimization of validation set 14 3 30 12 1 5
In order to provide a thorough assessment of the predic- 15 5 30 4 7 2
tive abilities, the validation set must include all concen- 16 5 10 10 1 4
tration ranges predicted in the calibration procedure. 17 1 25 4 5 5
Simple random sampling carries the risk of incomplete 18 4 10 8 7 5
coverage, which could result in prejudiced accuracy esti- 19 1 20 10 7 3
mates. To address this crucial constraint, the validation 20 3 25 10 3 2
set was carefully created using the KSC. To create a solid 21 4 25 6 1 3
and representative validation set that covers the whole 22 4 15 4 3 4
multivariate concentration space, the KSC algorithm is 23 2 10 6 5 2
carefully used. MATLAB was used to run the KSC algo- 24 1 15 8 1 2
rithm with the aid of pre-written scripts. The multicom- 25 2 20 4 1 6
ponent concentration data was methodically stratified Validation set (µg/mL)
into discrete clusters along several dimensions by these 1 3 15 11 2 4
scripts. One validation sample was chosen from each 2 3 23 8 4 3
stratum, which corresponds to a concentration regime 3 2 13 8 3 2
that is equally likely. This approach ensured comprehen- 4 4 13 6 8 4
sive and equitable coverage of the full analyte concen- 5 3 26 10 9 6
tration ranges and their combinatorial distributions. By 6 1 21 4 2 4
optimally distributing validation samples across clustered 7 2 29 7 5 4
data strata, KSC mitigated potential biases arising from 8 2 17 5 5 3
incomplete coverage or uneven sample densities within 9 4 10 6 6 5
the multivariate space. This meticulous, statistically 10 4 24 11 7 3
grounded sampling technique enabled a relatively small, 11 2 28 9 8 4
but very informative, validation set of 13 mixtures that 12 5 18 8 2 5
could rigorously test the predictive power of the chem- 13 4 21 12 6 5
ometric models over the whole range of expected sam-
ple compositions, as presented in (Table 1). The scatter
plots in (Figs. 4, 5) show how a consistent scattering pat-
smaller but more informative validation set to be applied.
tern was formed by the thirteen validation samples over
This tactic decreased the number of materials used, the
the whole analyte range. The KSC algorithm acquired
amount of waste produced, and the related expenses,
broad coverage with markedly fewer samples than stand-
which enhanced the method’s greenness. Furthermore,
ard random sampling. Through improved concentration
the model’s resilience in managing diverse combinations
space sampling efficiency, the KSC algorithm allowed a
Abbas et al. BMC Chemistry (2024) 18:233 Page 10 of 21

Fig. 4 2D scatter plot of the validation set of (a) CAF versus CAM, b CAF versus MEC, c ERG versus CAF, d ERG versus CAM, e ERG versus MEC, and f
CAM versus MEC designed by KSC design as optimal-space filling design
Abbas et al. BMC Chemistry (2024) 18:233 Page 11 of 21

Fig. 5 2D scatter plot of the validation set of (a) ERG versus PRO, b PRO versus CAM, c PRO versus MEC, and d PRO versus CAF designed by KSC
design as optimal-space filling design

of pharmaceutical compounds was shown by the precise initial predictions were inadequate. However, incorpo-
predictions made on this meticulously prepared KSC rating an intercept term greatly enhanced results, result-
validation set. This successfully avoided the erroneous or ing in exceptional recovery percentages of 99.65, 100.14,
excessively optimistic accuracy estimations that can arise 100.89, 100.41, and 99.38% for ERG, PRO, CAF, CAM
from insufficient sample variety and coverage in the vali- and MEC respectively. While simple, CLS has trouble
dation set. with interactions and nonlinearity. All things considered,
it offered a fair foundation for analysis, eventually being
CLS model exceeded by more advanced chemometric models.
The multivariate linear regression employed by the CLS
model, referred to as K-matrix calibration, is founded PLS model
on the Beer-Lambert law. Calibration samples require The PLS model is an additional popular chemomet-
analyte concentrations and spectral characteristics [24]. ric model that integrates PCA and regression analy-
CLS makes the assumption that absorbance and concen- sis. It optimizes the correlation amidst concentrations
tration for each component have a linear relationship. as response variables and spectra as predictor vari-
Though the calibration set was used to build CLS models, ables. This retains elements that are significantly more
Abbas et al. BMC Chemistry (2024) 18:233 Page 12 of 21

associated with concentration prediction. PLS steadily MCR-ALS iterations progressed, it became evident that
creates these latent variables by removing asymmetrical a five-component model was necessary to accurately
spectral data which is useless for estimating concentra- represent all analytes in the formulation. We applied
tions. We used leave-one-out cross-validation to ascer- non-negativity constraints to both spectral profiles and
tain the optimal number of PLS factors [13]. As shown in concentration, ensuring physically meaningful results,
(Fig. 6), five LVs in this investigation had the best mode- and implemented correlation constraints for the con-
ling results for ERG, PRO, CAF, CAM, and MEC, accom- centration profiles to account for potential interfer-
panied by RMSECV values of 0.062, 0.330, 0.129, 0.026, ences [26]. The MCR-ALS algorithm converged after
and 0.019. 15 iterations, achieving the predetermined convergence
criterion of 20%.
MCR‑ALS model To validate the accuracy of the spectral recovery, we
The MCR is a powerful chemometric model designed to performed a Cosine Similarity Analysis between the
elucidate the underlying spectral and concentration char- MCR-ALS resolved spectra and the independently
acteristics of distinct components within intricate combi- measured pure component spectra. Cosine similar-
nations, even in the absence of a priori information. The ity provides a quantitative measure of how closely the
MCR approach employs a bilinear model to decompose recovered spectra match the true spectra, with values
the spectral data matrix, yielding significant informa- approaching 1 indicating a near-perfect match. The
tion regarding the sample’s chemical composition. In our cosine similarity is defined as:
study, we implemented the MCR-ALS algorithm, which   
proceeds through several key steps. cos(θ ) = (A · B) / A ||B||
Initially, we employed Evolving Factor Analysis (EFA)
where A and B are the vectors representing the recovered
with a log eigenvalue threshold of -4 to estimate the
and standard spectra, respectively.
number of components present in the mixture. This
This analysis revealed excellent recovery for most
was followed by an Alternating Least Squares (ALS)
components, with cosine similarity values exceeding
optimization process, which iteratively refines the con-
0.99 for PRO (0.998), MEC (0.997), ERG (0.996), and
centration and spectral profiles subject to user-defined
CAF (0.995). A slightly lower, yet still acceptable, simi-
constraints. The process continues until convergence,
larity value was observed for CAM (0.982), indicating
defined by a relative change in the standard devia-
some minor deviations in its recovered spectral profile,
tion of residuals falling below a predefined threshold.
likely due to spectral overlap with other components.
Upon convergence, the algorithm produces three key
This comprehensive similarity assessment provides
matrices: the resolved spectral profiles, concentra-
quantitative validation of the model’s spectral resolu-
tion profiles, and associated error matrix [10, 25]. In
tion capabilities, with an average cosine similarity of
our implementation, we initially employed a three-
0.994 (standard deviation: 0.006) across all compo-
factor model based on the EFA results. However, as the
nents, as presented in (Table S2).
The model demonstrated excellent performance, with
a variance explained (­R2) of 100% and a remarkably low
20 lack of fit (0.0057%), indicating a high degree of agree-
ment between the model and experimental data. (Fig. 7)
18
presents the resolved pure spectral profiles obtained
16
ERG from the MCR-ALS analysis. The profiles show a high
14 PRO level of concordance with the measured absorption spec-
12 CAF tra of the individual components, validating the model’s
RMSECV

10 CAM ability to accurately deconvolute the overlapping spectral

8 MEC data. The application of MCR-ALS provided both quan-
titative precision and qualitative insights into the sam-
6
ple, demonstrating its dual capability in spectroscopic
4
analysis. This capability is particularly useful for resolving
2 complex formulations like the one studied, where over-
0 lapping spectra from multiple components create signifi-
1 3 5 7 9 11
cant challenges for analysis.
Latent variable
To further investigate the uncertainty inherent in
Fig. 6 RMSECV plot of the calibration set as a function
the MCR-ALS model, a Comprehensive Rotational
of the optimum LVs for the PLS model
Abbas et al. BMC Chemistry (2024) 18:233 Page 13 of 21

0.500 0.500

PRO absorption spectrum

CAM absorption spectrum

0.400 PRO resolved spectrum 0.400

CAM resolved spectrum

0.300 0.300

0.200 0.200

0.100 0.100

0.000 0.000
220.00 250.00 300.00 350.00 220.00 250.00 300.00 350.00
Wavelength (nm) Wavelength (nm)

0.500 0.250

MEC absorption spectrum

CAF absorption spectrum
0.400 0.200
MEC resolved spectrum
CAF resolved spectrum

0.300 0.150

0.200 0.100

0.100 0.050

0.000
0.000
220.00 250.00 300.00 350.00
220.00 250.00 300.00 350.00
Wavelength (nm)
Wavelength (nm)

0.125

ERG absorption spectrum

0.100
ERG resolved spectrum

0.050

0.000
220.00 250.00 300.00 350.00
Wavelength (nm)

Fig. 7 Absorption spectra and resolved spectra estimated by MCR-ALS algorithm

Ambiguity Assessment was conducted using the MCR- 0.024 AU at a critical wavelength of 278 ± 2 nm, result-
BANDS algorithm. This assessment quantified the Area ing in minimal impact on quantitation (< 2.1%). PRO
of Feasible Solutions (AFS), the Average Range of Fea- exhibited the lowest ambiguity, with an AFS of 4.5%, an
sible Solutions (Avg RFS), the Maximum Range of Fea- average RFS of 0.008 AU, and a maximum RFS of 0.018
sible Solutions (Max RFS), and identified the critical AU at 242 ± 1 nm, with negligible impact on quantita-
wavelengths where ambiguity was most pronounced. tion (< 1.4%). CAF had moderate ambiguity, with an
As shown in (Table S3), the AFS for ERG was 7.2%, with AFS of 5.8%, an average RFS of 0.009 AU, and a maxi-
an average RFS of 0.011 AU and a maximum RFS of mum RFS of 0.021 AU at 273 ± 2 nm, resulting in a
Abbas et al. BMC Chemistry (2024) 18:233 Page 14 of 21

minor impact on quantitation (< 1.8%). CAM showed requires careful consideration due to its higher ambigu-
the highest ambiguity, with an AFS of 8.9%, an aver- ity, which has been factored into the final concentration
age RFS of 0.015 AU, and a maximum RFS of 0.029 AU determinations. These results demonstrate the robust-
at 265 ± 3 nm, leading to a moderate impact on quan- ness and versatility of the MCR-ALS model in resolving
titation (< 3.2%). MEC exhibited moderate ambiguity, complex mixtures, ensuring both accurate quantifica-
with an AFS of 6.3%, an average RFS of 0.010 AU, and a tion and valuable qualitative insights.
maximum RFS of 0.022 AU at 260 ± 2 nm, with a minor
impact on quantitation (< 1.9%). The overall mean AFS Validation of the chemometric models
across all components was 6.54%, with a standard devi- The assessment of predictive abilities was conducted
ation of 1.68%, indicating moderate ambiguity across utilizing an external validation set comprising thirteen
the components. The correlation coefficient between mixtures produced from KSC. These validation mixtures
AFS and quantitation impact was 0.92, suggesting a were not integrated into the model construction process,
strong correlation between rotational ambiguity and its though their concentration levels remained within the
effect on quantitative results. Detailed analysis of each working range employed to develop the models, as dem-
component revealed that despite moderate ambiguity onstrated in (Table 1). To assess their predictive power,
for ERG, its well-defined spectral features minimized the newly constructed analytical models were applied to
the impact on quantitation, with a selectivity index of these validation mixtures, with the estimated concentra-
0.89. PRO showed the lowest ambiguity and highest tions for each component presented in (Table 1).
selectivity index (0.94), thanks to its higher concentra- To statistically evaluate the accuracy of the predictions,
tion and stable spectral features. CAF also had moder- Student’s t-test was performed for each analyte across all
ate ambiguity, but its distinct spectral profile ensured three models (CLS, PLS, and MCR-ALS) to determine
reliable resolution, with a selectivity index of 0.91. if the mean recoveries were significantly different from
CAM displayed the highest ambiguity due to signifi- 100% [27]. The critical t values at 95% and 98% confi-
cant spectral overlap, reflected in its selectivity index dence levels ­(df = 12) were 2.179 and 3.055, respectively.
of 0.85. This suggests that additional constraints may For the CLS model, experimental t values ranged from
be beneficial for further reducing ambiguity for CAM. 0.522 to 2.359, with only PRO showing a significant dif-
MEC was well-resolved from other components, with ference from 100% ­(texp = 2.359). The PLS model exhib-
moderate ambiguity and a selectivity index of 0.90. ited similar results, with experimental t values between
To mitigate the impact of ambiguity, correlation con- 0.174 and 2.349, again with PRO being significantly dif-
straints were employed, reducing the AFS by an average ferent from 100% ­ (texp = 2.349). The MCR-ALS model
of 42%, while non-negativity constraints provided addi- demonstrated the most consistent results, with experi-
tional stability. Local rank analysis further improved mental t values ranging from 0.334 to 1.980, indicating
the resolution, particularly for CAM and MEC, where no significant differences from 100% for any analyte at
spectral overlap was more pronounced. The overall AFS both confidence levels, as shown in (Table S4).
values, all below 10%, indicate good stability and reli- The root mean square error of prediction (RMSEP) val-
able quantitation for all components. However, CAM ues were determined to be minimal, with the MCR-ALS

Table 2 Determination of ERG, PRO, CAF, CAM, and MEC in the calibration and validation set of the suggested methods
CLS PLS MCR-ALS
ERG PRO CAF CAM MEC ERG PRO CAF CAM MEC ERG PRO CAF CAM MEC

Calibration set MEAN 99.65 100.14 100.89 100.41 99.38 99.69 99.99 100.88 100.12 99.49 99.96 99.86 100.32 99.94 99.56
SD 1.550 1.522 1.110 1.566 1.414 1.293 1.340 1.117 1.385 1.272 0.679 1.083 0.837 0.807 0.823
%RSD 1.555 1.520 1.100 1.560 1.423 1.297 1.340 1.107 1.383 1.279 0.679 1.085 0.834 0.807 0.827
RMSECa 0.087 0.456 0.231 0.178 0.104 0.079 0.415 0.210 0.162 0.094 0.072 0.378 0.191 0.147 0.086
Validation set MEAN 100.12 100.64 100.82 99.79 100.32 100.07 100.41 100.82 99.94 100.10 100.10 100.14 100.31 99.89 99.99
SD 0.517 1.472 1.255 1.452 0.78 0.310 1.143 1.260 1.243 0.370 0.182 0.526 0.603 0.664 0.108
%RSD 0.516 1.463 1.245 1.455 0.778 0.310 1.138 1.250 1.244 0.370 0.182 0.525 0.601 0.665 0.108
RMSEPb 0.093 0.487 0.249 0.189 0.112 0.085 0.443 0.226 0.172 0.102 0.077 0.404 0.205 0.156 0.093
a
Root Mean Square Error of calibration
b
Root Mean Square Error of predication
Abbas et al. BMC Chemistry (2024) 18:233 Page 15 of 21

model demonstrates superior results, as evidenced by Table 3 Determination of ERG, PRO, CAF, CAM, and MEC in
RMSEP values of 0.077, 0.404, 0.205, 0.156, and 0.093 for pharmaceutical preparation by the suggested chemometric
ERG, PRO, CAF, CAM, and MEC, respectively (Table 2). methods and application of standard addition technique
The RRMSEP offered a metric for prediction accuracy Preparation %Recovery ± %RSDa
represented as a proportion of the mean concentration
CLS PLS MCR-ALS
of the analyte. The MCR-ALS model revealed favorable
RRMSEP values of 3.121, 1.936, 2.487, 2.936, and 2.386% ERG Application 99.22 ± 0.673 99.68 ± 0.491 99.95 ± 0.261
for ERG, PRO, CAF, CAM, and MEC respectively, as Standard addition 100.69 ± 0.825 100.86 ± 0.603 100.98 ± 0.381
shown in (Table S4). The BCMSEP supplied informa- PRO Application 101.31 ± 0.957 101.17 ± 0.735 101.05 ± 0.513
tion about the precision and variability of the prediction, Standard addition 98.43 ± 0.926 99.31 ± 0.704 99.84 ± 0.482
with all models and components achieving acceptably CAF Application 98.11 ± 0.963 99.13 ± 0.741 99.79 ± 0.519
low values, suggesting high precision in the predictions Standard addition 101.97 ± 0.994 101.63 ± 0.772 101.31 ± 0.550
(Table S4). To evaluate sensitivity, LOD and LOQ were CAM Application 101.21 ± 0.999 101.85 ± 0.777 101.51 ± 0.555
calculated from the net analyte signals, demonstrating Standard addition 99.22 ± 0.673 99.68 ± 0.491 99.95 ± 0.261
that the models were sufficiently sensitive for pharma- MEC Application 98.76 ± 0.892 99.48 ± 0.670 99.89 ± 0.448
ceutical examination. Furthermore, the models exhib- Standard addition 101.31 ± 0.957 101.17 ± 0.735 101.05 ± 0.513
ited remarkable precision and robustness upon repeated a
Average of three determinations
testing, with percent relative standard deviation (%RSD)
values below 2% for both intra-day and inter-day meas-
urements, as presented in (Table S4). hand, makes the assumption that response variables and
spectral data have a linear connection offers greater ver-
Statistical analysis
satility by efficiently determining significant components
There was no discernible difference in accuracy across even in the presence of unknown elements. However,
the different models that were suggested, according to MCR-ALS exhibited exceptional performance due to its
the One-way analysis of variance (ANOVA) conducted unique ability to iteratively learn system behavior from
on the validation dataset. As shown in (Table S5), the dynamic data, achieving a comprehensive mathematical
p-values were more than 0.05 The calculated F-values and chemical understanding of system complexity. Nota-
were inferior to the critical F-value, indicating that there bly, MCR-ALS excelled in resolving pure spectra of indi-
were no statistically substantial variations in the accuracy vidual components, enabling qualitative identification of
of these models. Furthermore, the chemometric mod- impurities, so strengthening its resilience to unidentified
els proposed in this research exhibited comparable per- interferences.
formance to the previously published approach [28] for To further validate these models, the Elliptical Joint
the determination of ERG, PRO, CAF, CAM and MEC Confidence Region (EJCR) test was employed, graphi-
concentrations, as evidenced by the data presented in cally assessing their predictive performance [27, 29]. The
(Table S5). EJCR plot (Fig. S2) provide insightful information about
how well the three models performed in comparison
across five distinct medications. MCR-ALS consistently
Pharmaceuticals assay
demonstrated superior performance, with its confidence
SPASMOMIGRAINE® tablets without interfering with
The recommended method worked well for testing
ellipses being the smallest and closest to the ideal point
(slope = 1, intercept = 0) for the majority of analytes. This
the additives. The recommended models were validated
indicates the best alignment between true and predicted
using the technique of standard addition, and the out-
values and the highest accuracy among the models. PLS
comes are displayed in (Table 3).
showed variable performance, excelling for some drugs
(e.g., CAF and MEC) with ellipses comparable to MCR-
Comparative evaluation of chemometric models ALS, but performing poorly for others (e.g., ERG and
The comparative analysis of chemometric models, CAF) with larger, more offset ellipses. CLS generally
including PLS, CLS, and MCR-ALS, revealed MCR-ALS exhibited the largest ellipses and greatest deviations from
as the superior model across multiple performance met- the ideal point, underscoring its limitations in handling
rics. This conclusion was drawn based on key validation complex mixtures.
parameters such as RMSCP, RMSEC, SD, R% values, as Drug-specific analysis revealed interesting patterns.
detailed in (Table S4). While CLS demonstrated adequate For ERG, MCR-ALS significantly outperformed both PLS
performance, its reliance on precise knowledge of all cali- and CLS, while all models struggled with PRO, suggest-
bration sample components limits its practical applica- ing it may be particularly challenging to analyze. CAF
bility, especially in complex mixtures. PLS, on the other
Abbas et al. BMC Chemistry (2024) 18:233 Page 16 of 21

and MEC showed closer competition between MCR- varies from 2 to 12. Furthermore, waste generation is
ALS and PLS, with both models performing well. CAM below 50 g (Fig. S3). NEMI pictograms were created for
displayed the most variability among models, with PLS the suggested methodology (Table 4). An early study of
unexpectedly showing the largest ellipse, while MCR- the pictograms revealed that the suggested technique sat-
ALS maintained its superior performance. isfied all NEMI criteria, as evidenced by the four green
The shape and orientation of the ellipses provided quadrants.
additional insights. Most ellipses were elongated diago-
nally, indicating a correlation between slope and inter-
cept errors. The tendency for ellipses to have slopes Complex GAPI tool
slightly greater than 1 and negative intercepts suggests While NEMI provides a preliminary assessment of green-
a general trend across models to slightly overestimate ness, the Complex GAPI tool delivers a more thorough
at higher concentrations and underestimate at lower semi-quantitative analysis [36]. It surpasses the authen-
concentrations. tic GAPI measure by incorporating a hexagonal area that
In conclusion, this comprehensive comparative study, delineates the phases and stages preceding the analysis
encompassing both traditional validation parameters and (Fig. S4). This cutting-edge tool includes every step of
the EJCR analysis, unequivocally demonstrates the supe- the analytical process, from getting samples and moving
rior predictive accuracy and consistency of MCR-ALS, them to keeping them safe, storing them, preparing them,
particularly in complex chemical systems. PLS emerges and finally analyzing them [36]. Furthermore, Complex
as a viable alternative, especially for specific analytes GAPI provides user-friendly software for the creation of
where it performs comparably to MCR-ALS. CLS, while optical pictograms. In this research, the offered meth-
functional, shows significant limitations in accuracy and odology demonstrated significant benefits, as evidenced
precision compared to the other models, especially for by a minor E-factor of 1 and a prevalence of satisfac-
complex or challenging analytes. These findings under- tory green quadrants, signifying low waste production
score the importance of model selection based on the and a favorable influence on sustainability, as revealed
specific characteristics of the analytes being studied and in (Table 4). Environmental considerations are the main
highlight MCR-ALS as the most reliable choice for multi- focus of Complex GAPI; however, incorporating com-
component analytical scenarios. plementary quantitative measures with complex GAPI
facilitates a more thorough assessment that includes
Greenness, blueness, and whiteness appraisal multi-dimensional sustainability metrics such as energy
It is essential to evaluate the environmental and eco- preservation, waste reduction, and the utilization of reju-
nomic impacts of analytical processes to comprehend venated resources.
their sustainability. Sustainability is a multi-sided notion
that includes elements like environmental friendli-
ness, waste reduction, safety, effectiveness, and afford- AGREE tool.
ability [30–32]. A thorough assessment of sustainability The AGREE metric, encompassing all 12 GAC concepts,
across all important factors cannot be achieved with a offers a valuable quantitative approach for assessing
single tool [33–35]. For this reason, this study employs greenness [37]. This facilitates a comprehensive evalu-
a thorough multiple-tool approach to facilitate a more ation based on commonly acknowledged GAC criteria.
thorough evaluation from multiple supplementary One of AGREE’s main advantages is the flexibility it pro-
viewpoints. vides through the adjustable weighting of these various
criteria. The user-friendly software converts the 12 inputs
NEMI tool into a singular score ranging from 0 to 1, presenting it on
The NEMI study conducted a preliminary assessment a vibrant pictogram for rapid comprehension. Dark red
of greenness inadequacies using an optical representa- denotes significant shortcomings, whereas dark green
tion divided into four quadrants [13]. The four quadrants denotes superior greenness. In this investigation, the
encompass one PBT (persistent, bioaccumulative, and suggested approach received a superior AGREE score of
toxic) substance, two hazardous materials, three corro- 0.75 (Table 4), confirming its exceptional effectiveness in
sives, and four waste products. The green quadrant signi- furthering sustainability objectives. On the other hand,
fies the following: (1) The reagents used are not classified AGREE only considers environmental factors. Integrat-
as PBT by the Toxic Release Inventory (EPA-TRI) of the ing AGREE with tools that evaluate other important fac-
Environmental Protection Agency; (2) the compounds tors like affordability, practicality, safety, and analytical
utilized are deemed non-hazardous and are consequently ability is helpful in enabling a comprehensive sustainabil-
omitted from the TRI list; and (3) the pH of the medium ity evaluation.
Abbas et al. BMC Chemistry (2024) 18:233 Page 17 of 21

Table 4 Comparative study of the proposed and reported methods

The proposed method The reported method (Elbarbry et al., 2007)

NEMI tool

ComplexGAPI tool

AGREE tool

Carbon footprint (kg CO2 eq

/sample)
0.
09
7
0.002

Proposed method

Reported method
Abbas et al. BMC Chemistry (2024) 18:233 Page 18 of 21

Table 4 (continued)
The proposed method The reported method (Elbarbry et al., 2007)

BAGI tool

RGB12 tool R1: Scope of G1: Toxicity B1: Cost-

R1: Scope of
90.0
G1: Toxicity
100.0
B1: Cost-
90.0 applicaon 50.0 of reagents 60.0 efficiency 90.0
applicaon of reagents efficiency
G2: Amount
G2: Amount R2: LOD and B2: Time-
R2: LOD and B2: Time- 60.0 of reagents 50.0 60.0
LOQ 85.0 of reagents 80.0 efficiency 90.0 LOQ
and waste
efficiency
and waste
G3: Energy
G3: Energy B3:
B3: R3: Precision 80.0 and other 50.0 75.0
R3: Precision 90.0 and other 90.0 Requirements 92.5 media
Requirements
media
G4: Direct B4: Operaonal
G4: Direct B4: Operaonal R4: Accuracy 85.0 86.7 66.7
R4: Accuracy 95.0 impacts 100.0 simplicity 75.0 impacts simplicity

90.0 92.5 86.9 68.8 61.7 72.9

89.8 67.8

Carbon footprint analysis less of a carbon footprint because it uses less electricity
Carbon footprint analysis, as opposed to other greenness because the analysis times are shorter and there is no
evaluations, allows for a quantitative comparison of the derivatization step. Moreover, using ethanol instead of
ecological effects of analytical methods with respect to dangerous solvents like chloroform and methylene chlo-
greenhouse gas emissions, quantified in kilograms of ­CO2 ride greatly reduced emissions related to transportation,
equivalent [38]. While tools such as AGREE, Complex proving that it is good for the environment.
GAPI, and NEMI provide valuable insights into method
greenness, they lack the capability to quantify emissions BAGI tool
directly. Carbon footprint analysis, however, presents a BAGI represents a paradigm shift in analytical method
comprehensive measure that incorporates crucial fac- evaluation, moving beyond the traditional focus on envi-
tors often overlooked in other assessments. These fac- ronmental "greenness" to quantify a method’s practical
tors include power consumption, reagent transportation, applicability or "blueness" [40]. This innovative metric
and waste generation – all significant contributors to the provides a comprehensive assessment of an analytical
overall environmental impact of analytical procedures. procedure’s real-world viability by examining ten critical
To conduct our carbon footprint analysis, we employed operational parameters. These parameters encompass the
a standardized equation [39]: nature of the analysis, analyte multiplicity, instrumental

Carbon footprint (kg CO2 eq) = Instrument Power(kW ).Analysis time(h).Emission factor (kgCO2/kWh)

The suggested approach had a substantially minimal

requirements, sample throughput efficiency, preparation
carbon footprint, merely 0.002-kg C ­ O2 equivalent for
complexity, hourly analysis capacity, resource demands,
each sample, as revealed in (Table 4). Our method leaves
Abbas et al. BMC Chemistry (2024) 18:233 Page 19 of 21

pre-concentration necessities, automation extent, and other metrics allowed for a thorough, reliable assess-
sample volume considerations. The BAGI methodol- ment of sustainability and got around the drawbacks of
ogy assigns each factor a score on a scale of 1 to 10, with using just one method. An objective, thorough analyti-
higher scores indicating superior performance. By cal- cal approach to sustainability assessment is best dem-
culating the geometric mean of these individual scores, onstrated by this systems-oriented mindset that makes
the BAGI generates a single, holistic value that reflects use of numerous complementary tools.
the method’s overall suitability for routine implementa-
tion. This approach yields valuable insights into a meth-
od’s relevance, functionality, and fitness-for-purpose in Conclusion
practical settings, offering a nuanced perspective that This innovative study effectively developed and validated
complements environmental considerations. Conse- three chemometric models (CLS, PLS, and MCR-ALS)
quently, analytical procedures boasting higher BAGI for the concurrent environmentally friendly determi-
scores are better positioned to meet the multifaceted nation of ERG, PRO, CAF, CAM and MEC using UV–
demands of real-world applications, balancing ecological visible spectroscopy. The deliberate use of the KSC
responsibility with operational efficacy. In this study, our guaranteed strong validation across various multivariate
method attained an impressive BAGI score of 90, signify- concentration levels. The rigorous evaluation confirmed
ing exceptional applicability. According to (Table 4), the the precision, accuracy, and sensitivity of these models,
BAGI assessment approves that our method has a great with MCR-ALS emerging as the top performer by quali-
deal of benefits with respect to hazard reduction, time tatively resolving pure component spectra in addition to
and money savings, and general usability. BAGI does not, quantification. The embrace of Principles of green and
however, offer a comprehensive, all-encompassing quan- white analytical chemistry, coupled with the compre-
tification of sustainability, even though it assesses impor- hensive greenness, blueness, and whiteness assessment
tant real-world use. To get a more all-encompassing using six cutting-edge tools, positions this method as a
evaluation that takes environmental friendliness, analyti- sustainable, practical, and analytically powerful alterna-
cal quality, and practicality into account, we also used the tive to conventional chromatographic techniques. The
RGB12 tool. eco-friendly solvent selection, reduced waste, low car-
bon footprint, and high analytical greenness reinforce its
RGB12 tool potential for widespread implementation in cost-effective
In 2021, Paweł Nowak and his team publicly introduced routine quality control. This novel methodology ena-
the RGB12 algorithm [41]. It is an easy-to-use quan- bles the on-site measurement of ERG, PRO, CAF, CAM,
titative assessment instrument for evaluating white- and MEC drugs using affordable UV instrumentation,
ness. This instrument evaluates methodologies based enhancing pharmaceutical quality assurance in resource-
on the 12 WAC issues and examines their sustainabil- limited settings.
ity in relation to whiteness appraisal [42]. The RGB12
algorithm consists of twelve unique algorithms, catego- Supplementary Information
The online version contains supplementary material available at https://​doi.​
rized into three groups: "red," "green," and "blue," with org/​10.​1186/​s13065-​024-​01339-4.
each category containing four methods. The green sub-
group (G1–G4) examines critical GAC factors includ- Additional file 1
ing toxicity, energy conservation, reagent production,
and waste and effects on animals, humans, and genetic Acknowledgements
alterations. The red group (R1–R4) focuses on valida- The authors extend their appreciation to Taif University, Saudi Arabia, for sup-
tion criteria such as accuracy, limit of quantification porting this work through the project number (TU-DSPP-2024-49).
(LOQ), precision, application scope, and limit of detec- Author contributions
tion (LOD). The blue subgroup (B1–B4) relates to prac- A.E.F.A. Methodology, design of the work, investigation, writing-original draft,
tical and economic requirements, cost-effectiveness, and writing review and editing. N.A.A. and M.G. Supervision, and investiga-
tion. M.K.H., B.A.M.S. and I.A.N Writing review and editing. M.M.A.M and S.M.M.
and time efficiency. The RGB12 algorithm adheres to Interpretation of data and figures preparation. Y.A.S: Supervised analysis proce-
WAC tenets by consolidating the scores across all three dures and carried out sample preparation. All the authors read, reviewed, and
color domains to calculate the final "whiteness" number. approved the manuscript.
The approach attains a notable whiteness score of 89.8, Funding
as indicated in (Table 4). This proves that the method This research was funded by Taif University, Saudi Arabia, Project No.
offers several usefulness in terms of being eco-friendly, (TU-DSPP-2024–49).
sustainable, economically viable, practically applica-
ble, and analytically efficient. Combining RGB12 with
Abbas et al. BMC Chemistry (2024) 18:233 Page 20 of 21

Availability of data and materials spectrophotometry and chemometric tools. Spectrochim Acta A Mol
Data was collected using a spectrophotometer and software. The correspond- Biomol Spectrosc. 2020. https://​doi.​org/​10.​1016/j.​saa.​2020.​118415.
ing author will provide the datasets created and/or analyzed during the 9. Attia KAM, El-Olemy A, Abbas AEF, Eid SM. A sustainable data processing
current study upon reasonable request. approach using ultraviolet-spectroscopy as a powerful spectral resolution
tool for simultaneously estimating newly approved eye solution in the
presence of extremely carcinogenic impurity aided with various green-
Declarations ness and whiteness assessment perspectives: Application to aqueous
humor. J Chem Res. 2023;47:5.
Ethics approval and consent to participate 10. Rahman MAA, Elghobashy MR, Zaazaa HE, El-Mosallamy SS. Novel
Not applicable. analytical method based on chemometric models applied to UV–Vis
spectrophotometric data for simultaneous determination of Etoricoxib
Consent for publication: and paracetamol in presence of paracetamol impurities. BMC Chem.
Not applicable. 2023;17:176.
11. Taha AM, Elmasry MS, Hassan WS, Sayed RA. Spider chart, greenness and
Competing interests whiteness assessment of experimentally designed multivariate models
The authors declare no competing interests. for simultaneous determination of three drugs used as a combinatory
antibiotic regimen in critical care units: comparative study. Spectrochim
Author details Acta A Mol Biomol Spectrosc. 2024;313: 124115.
1
Faculty of Pharmacy, Analytical Chemistry Department, October 6 University, 12. Sami I, Rostom Y, Yehia AM, El-Saharty YS, Monir HH. Environmentally safe
6 October City 12585, Giza, Egypt. 2 Department of Pharmaceutical Analytical chemometric evaluation and data analysis for the kinetic investigation of
Chemistry, School of Pharmacy, Badr University in Cairo, Badr City 11829, Cairo, mirabegron stability. Sustain Chem Pharm. 2024;39: 101594.
Egypt. 3 Pharmaceutical Analytical Chemistry Department, Faculty of Phar- 13. Attia KAM, El-Olemy A, Eid SM, Abbas AEF. A green-and-white integrative
macy, Beni-Suef University, Alshaheed Shehata Ahmad Hegazy St, Beni‑Suef, analytical strategy combining univariate and chemometric techniques
Egypt. 4 College of Pharmacy, Al-Mustaqbal University, Babylon 51001, Iraq. for quantifying recently approved multi-drug eye solution and potentially
5
Department of Pharmaceutical Chemistry, College of Pharmacy, Taif Univer- cancer-causing impurities: application to the aqueous humor. J AOAC Int.
sity, Taif, Saudi Arabia. 6 Chemistry Department, Faculty of Science, University 2024;107:146–57.
of Al-Jufra, P.O. Box 61602, Al‑Jufra, Libya. 7 Pharmaceutical Analytical Chemistry 14. Eid SM, Hassan SA, Nashat NW, Elghobashy MR, Abbas SS, Moustafa
Department, Faculty of Pharmacy, Damanhour University, Damanhour, Egypt. AA. Optimization of localized surface plasmon resonance hot spots in
8
Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Sinai Univer- surface-enhanced infrared absorption spectroscopy aluminum substrate
sity -Kantara Branch, Ismailia 341636, Egypt. as an optical sensor coupled to chemometric tools for the purity assay of
quinary mixtures. Microchim Acta. 2021;188:195.
Received: 30 August 2024 Accepted: 30 October 2024 15. Li T, Fong S, Wu Y, Tallon-Ballesteros AJ. Kennard-Stone balance algorithm
for time-series big data stream mining. In: 2020 International Conference
on Data Mining Workshops (ICDMW). IEEE. 2020.
16. Brereton RG, Jansen J, Lopes J, Marini F, Pomerantsev A, Rodionova O,
et al. Chemometrics in analytical chemistry—part II: modeling, validation,
References and applications. Anal Bioanal Chem. 2018;410:6691–704.

components of spasmomigraine® tablet by liquid chromatography with

1. Elbarbry FA, Mabrouk MM, El-Dawy MA. Determination of the analgesic 17. Mostafa A, Shaaban H. Chemometric assisted uv-spectrophotometric
methods using multivariate curve resolution alternating least squares
ultraviolet detection. J AOAC Int. 2007;90:94–101. and partial least squares regression for determination of beta-antagonists
2. Elbordiny HS, Elonsy SM, Daabees HG, Belal TS. Design of trio-colored vali- in formulated products: evaluation of the ecological impact. Molecules.
dated HPLC method for synchronized multianalyte quantitation of four 2022;28:328.
top selling antihyperlipidemic drugs in different fixed-dose combined 18. Larsen C, Lundberg P, Tang S, Ràfols-Ribé J, Sandström A, Mattias Lindh
tablets. Green Analyt Chem. 2024;8: 100100. E, et al. A tool for identifying green solvents for printed electronics. Nat
3. Saleh SS, Obaydo RH, El Hamd MA, Rostom Y, Mohamed D, Lotfy HM. Commun. 2021;12:4510.
Guidelines for accurate application of green and white analytical 19. Kayali Z, Obaydo RH, Alhaj SA. Spider diagram and sustainability evalu-
concepts: Merits Versus demerits with insights of significant milestones ation of UV-methods strategy for quantification of aspirin and sildenafil
of assessment tools applied for antiviral drugs. Microchem J. 2024;199: citrate in the presence of salicylic acid in their bulk and formulation.
109917. Heliyon. 2023. https://​doi.​org/​10.​1016/j.​heliy​on.​2023.​e15260.
4. Korany MA, Youssef RM, Ragab MAA, Amine MT, Afify MA. Development 20. Shen Y, Lo C, Nagaraj DR, Farinato R, Essenfeld A, Somasundaran P. Devel-
and validation of stability indicating HPLC-DAD assay of adapalene, opment of greenness index as an evaluation tool to assess reagents:
methylparaben and propylparaben in a cosmeceutical gel matrix with evaluation based on SDS (Safety Data Sheet) information. Miner Eng.
blueness, greenness and whiteness assessment via relative ranking: an 2016;94:1–9.
ab initio-sub finem juxtaposed non-traditional approaches. Microchem J. 21. Ferraro MCF, Castellano PM, Kaufman TS. Chemometric determination
2024;205: 111342. of amiloride hydrochloride, atenolol, hydrochlorothiazide and timolol
5. Katamesh NS, Abbas AEF, Mahmoud SA. Four chemometric models maleate in synthetic mixtures and pharmaceutical formulations. J Pharm
enhanced by Latin hypercube sampling design for quantification of anti- Biomed Anal. 2004;34:305–14.
COVID drugs: sustainability profiling through multiple greenness, carbon 22. Saad AS, Elzanfaly ES, Halim MK, Kelani KM. Comparing the predictability
footprint, blueness, and whiteness metrics. BMC Chem. 2024;18:54. of different chemometric models over UV-spectral data of isoxsuprine
6. Attimarad M, Chohan MS, Katharigatta Narayanaswamy V, Nair AB, and its toxic photothermal degradation products. Spectrochim Acta A
Sreeharsha N, Shafi S, et al. Mathematically processed UV spectroscopic Mol Biomol Spectrosc. 2019;219:444–9.
method for quantification of chlorthalidone and Azelnidipine in bulk and 23. Kelani KM, Shalaby AA, Elmaamly MY. Spectrophotometric and chemo-
formulation: evaluation of greenness and whiteness. J Spectros. 2022. metric methods for simultaneous determination of two anti-hypertensive
https://​doi.​org/​10.​1155/​2022/​49651​38. drugs in their combined dosageform. Pharm Anal Acta. 2015;6(321):2.
7. Alnemari RM, Abdelazim AH, Almalki AH, Alqahtani AS, Alaqel SI, Alsulami 24. Abdelazim AH, Shahin M, Abu-khadra AS. Application of different chemo-
FT, et al. Application of signal processing techniques for the spectro- metric assisted models for spectrophotometric quantitative analysis of
scopic analysis of dolutegravir and lamivudine: a comparative assessment velpatasvir and sofosbuvir. Spectrochim Acta A Mol Biomol Spectrosc.
and greenness appraisal. BMC Chem. 2024;18:129. 2021;252: 119540.
8. Elsonbaty A, Serag A, Abdulwahab S, Hassan WS, Eissa MS. Analysis of 25. Farid JF, Mostafa NM, Fayez YM, Essam HM, ElTanany BM. Chemometric
quinary therapy targeting multiple cardiovascular diseases using UV quality assessment of paracetamol and phenylephrine hydrochloride
with paracetamol impurities; comparative UV-spectrophotometric
Abbas et al. BMC Chemistry (2024) 18:233 Page 21 of 21

implementation of four predictive models. Spectrochim Acta A Mol Publisher’s Note

Biomol Spectrosc. 2022. https://​doi.​org/​10.​1016/j.​saa.​2021.​120308. Springer Nature remains neutral with regard to jurisdictional claims in pub-
26. Hassan SA, Nashat NW, Elghobashy MR, Abbas SS, Moustafa AA. lished maps and institutional affiliations.
Advanced chemometric methods as powerful tools for impurity profiling
of drug substances and drug products: application on bisoprolol and
perindopril binary mixture. Spectrochim Acta A Mol Biomol Spectrosc.
2022;267: 120576.
27. Olivieri AC. Practical guidelines for reporting results in single- and
multi-component analytical calibration: a tutorial. Anal Chim Acta.
2015;868:10–22.
28. Patel B, Patel S. A specific high-performance thin-layer chromatography
method validated for estimation of mometasone furoate and olopata-
dine hydrochloride. Sep Sci Plus. 2023. https://​doi.​org/​10.​1002/​sscp.​
20230​0032.
29. Mazivila SJ, Lombardi JM, Páscoa RNMJ, Bortolato SA, Leitão JMM, Esteves
da Silva JCG. Three-way calibration using PARAFAC and MCR-ALS with
previous synchronization of second-order chromatographic data through
a new functional alignment of pure vectors for the quantification in the
presence of retention time shifts in peak position and shape. Anal Chim
Acta. 2021;1146:98–108.
30. Radwan AS, Magdy G, Belal F, Hadad GM, Elkhoudary MM, Salim MM.
Carbon quantum dots as nanosensors for the spectrofluorimetric deter-
mination of meloxicam in biological fluids and dosage forms: Greenness
assessment and application to content uniformity testing. Sustain Chem
Pharm. 2024;41: 101691.
31. Othman WM, Al-Zoman NZ, Darwish IA, Almomen A, Saad SS, Abdal-
lah FF, et al. Development of an eco-friendly capillary electrophoresis
method for the simultaneous determination of piperacillin, tazobactam
and ibuprofen in plasma samples: application to a pharmacokinetic study
in rats. RSC Adv. 2024;14:23378–91.
32. Halim MK, Badran OM, Abbas AEF. Sustainable chemometric methods
boosted by Latin hypercube technique for quantifying the recently FDA-
approved combination of bupivacaine and meloxicam in the presence of
bupivacaine carcinogenic impurity: comprehensive greenness, blueness,
and whiteness assessments. Microchem J. 2024;200: 110276.
33. Gamal M, Naguib IA, Panda DS, Abdallah FF. Comparative study of four
greenness assessment tools for selection of greenest analytical method
for assay of hyoscine N-butyl bromide. Anal Methods. 2021;13:369–80.
34. El-Masry AA, Abbas AEF, Salem YA. A dual methodology employing ion-
pair chromatography and built-in UV spectrophotometry for quantifying
recently approved combination of mometasone and indacaterol in a
novel combined metered dose inhaler: assessing the greenness, carbon
footprint, blueness, and whiteness. BMC Chem. 2024;18:143.
35. Attia KAM, Serag A, Eid SM, Abbas AEF. A New chemometrically assisted
UV spectrophotometric method for simultaneous determination of
tamsulosin and dutasteride in their pharmaceutical mixture. J AOAC Int.
2022;105(6):1755.
36. Płotka-Wasylka J, Wojnowski W. Complementary green analytical proce-
dure index (ComplexGAPI) and software. Green Chem. 2021;23:8657–65.
37. Gałuszka A, Migaszewski Z, Namieśnik J. The 12 principles of green
analytical chemistry and the SIGNIFICANCE mnemonic of green analytical
practices. TrAC - Trends Analyt Chem. 2013;50:78–84.
38. Muthu. Carbon footprint case studies. Singapore: Springer Singapore;
2021.
39. Ballester-Caudet A, Campíns-Falcó P, Pérez B, Sancho R, Lorente M, Sastre
G, et al. A new tool for evaluating and/or selecting analytical methods:
summarizing the information in a hexagon. TrAC, Trends Anal Chem.
2019;118:538–47.
40. El Hamd MA, Soltan OM, Abdelrahman KS, Fouad A, Saleh SF, Obaydo
RH, et al. Roth’s switch-on fluoremetric probe for green tracking and
quantifying of 1.4-dihydropyridine medication: evaluation of greenness,
whiteness, and blueness. Sustain Chem Pharm. 2023;36: 101294.
41. Nowak PM, Wietecha-Posłuszny R, Pawliszyn J. White analytical chemistry:
an approach to reconcile the principles of green analytical chemistry and
functionality. TrAC, Trends Anal Chem. 2021;138: 116223.
42. Nowak PM, Kościelniak P. What color is your method? adaptation of the
RGB additive color model to analytical method evaluation. Anal Chem.
2019;91:10343–52.