High-throughput DoE synthesis of chalcones and pyrazolines for fluorescent sensing

Abstract

Pyrazolines are attractive scaffolds for fluorescent sensors due to their unique photophysical properties and the ease of their synthesis from chalcone precursors. While both chalcones and pyrazolines have been widely explored, a systematic optimization of their synthesis using design of experiments (DoE) has yet to be reported. Here, we apply a data-driven DoE strategy, evaluating multiple variables across more than 105 datapoints, to establish highly optimized conditions for the Claisen–Schmidt condensation of chalcones and their subsequent conversion to pyrazolines. High-throughput screening of the resulting library revealed pyrazoline P10 with unusual “turn on” fluorescence at 590 nm in the presence of Zn²⁺ only. This integrated DoE–HTS workflow demonstrates a powerful approach for accelerating sensor discovery.

---

Introduction

The integration of modern data-driven methodology into synthetic organic chemistry is streamlining synthetic method development.¹ Among these, design of experiments (DoE) enables the systematic identification of optimal parameters with a minimal number of experiments, offering a powerful alternative to the traditional one-factor-at-a-time (OFAT) strategy.² For instance, recent examples include the use of DoE in Heck–Suzuki reactions,³ B(OCH₂CF₃)₃-mediated direct amidation,⁴ and solvent selection for multicomponent reactions.⁵ In parallel, high-throughput screening (HTS) platforms are accelerating the discovery process, with improvements reported in drug development,⁶ enhanced catalyst screening,⁷ and solar cell optimisation.⁸ Both DoE and HTS approaches benefit from modular, easily accessible chemical scaffolds. Chalcones are privileged structures⁹ with a diverse range of pharmacological properties, including anti-cancer,¹⁰ anti-inflammatory,¹¹ and anti-infective properties.¹² Typically, chalcones are prepared via a Claisen–Schmidt condensation¹³ from commercially available aromatic ketone and aldehyde starting materials (Fig. 1A). The diverse range of inexpensive, commercially available starting materials enables large modular libraries of chalcones to be synthesized and screened. Interest in Claisen–Schmidt condensations has grown considerably over the past sixty years (Fig. 1B). Importantly, while useful in themselves, chalcones can also serve as valuable precursors for pyrazolines (blue in Fig. 1A), another privileged structure with multiple medicinal properties.¹⁴ Pyrazolines with nanomolar activities for cancer cell lines¹⁵ and COX inhibitory activity have been reported.¹⁶ Pyrazolines can be aromatised to pyrazoles¹⁷ (red in Fig. 1A), providing access to another highly attractive privileged structure.¹⁸ Despite the wealth of applications chalcones and pyrazolines provide, no DoE study has optimised access to these scaffolds. This study addresses this research gap, providing efficient access to chalcone and pyrazolines via DoE while exploring HTS platforms in pursuit of novel fluorescent sensors.

Fig. 1 Access to pyrazoline and pyrazoles via the chalcone scaffold (panel A). Claisen–Schmidt condensations reported between 1960 and 2025 via the Reaxys database (panel B).

Pyrazolines display unique fluorescence properties for the design of fluorescent sensors for toxic metals such as Zn²⁺, Cu²⁺, and Hg²⁺.¹⁹ The modular nature of the pyrazolines scaffold enables fine-tuning of photophysical properties, a useful feature for sensor development. The use of pyrazoline²⁰ and pyrazole²¹ sensors for detecting toxic metals in biological environments represents a growing area of interest. Group 12 sensors, which selectively detect and distinguish Zn²⁺, Cd²⁺, and Hg²⁺, are particularly challenging. Zinc, the second most abundant transition metal in the human body, is vital for life,²² cadmium is highly toxic and linked to numerous cancers,²³ and mercury displays potent neurotoxicity.²⁴ Few fluorescent sensors can differentiate between these metals at different fluorescence wavelengths (λ_em).²⁵ Simple pyrazoline A is a highly selective “turn on” sensor for Zn²⁺ and Cd²⁺ in MeCN (Fig. 2).²⁶ Further studies revealed pyrazole B with improved selectivity for Zn²⁺/Cd²⁺ (Fig. 2).²⁷ Pyrazoline C is a “turn on” sensor for Fe³⁺/Fe²⁺ with yellow λ_em.²⁸ Substitution of the pyrazoline N1 methyl for phenyl reversed the fluorescent response from “turn on” to “turn off” (D in Fig. 2).²⁸ The hydrazone functional group is commonly incorporated into multi-analyte fluorescent sensors.²⁹ A recent breakthrough reported first-in-class hydrazone-pyrazolines sensors distinguishing Zn²⁺ at 560 nm, Cd²⁺ at 510 nm, and Hg²⁺ at 460 nm λ_em in aqueous environments (E in Fig. 2).³⁰ Typical approaches to chalcones involve the use of polar protic solvents, such as ethanol and methanol, accounting for 50% and 20% of all Claisen–Schmidt condensations reported between 1960–2025 (see SI1 for literature survey). Researchers often use previous synthetic methodologies with little to no experimental optimisation reported. This study addresses this research need using a data-driven design of experiments (DoE) approach. A large-scale chalcone library (>1 g scale) enabled a second DoE study focusing on pyrazoline synthesis. A pyrazoline library was generated, screened across multiple analytes using a HTS platform. This workflow significantly accelerated the discovery process, providing a valuable case study on incorporating data-driven methodology in synthetic organic chemistry.

Fig. 2 Fluorescent sensors for toxic metals. This study incorporated design of experiments (DoE) and high throughput screening (HTS) for pyrazoline sensor development (inset).

Results and discussion

An LC-MS method to monitor the Claisen–Schmidt condensation reaction was established (Fig. 3). An internal standard (acetaminophen) was added to all reaction aliquots ensuring injection consistency, and external calibration curves were generated (see SI3 for full details). A typical chromatogram is displayed in Fig. 3 demonstrating the conversion of 2-acetylpyridine and benzaldehyde to C1. A screening study to confirm the suitability of the LC-MS method and determine the main DoE parameters was performed (SI4). Solvent, consisting of the bulk of the reaction mixture, had a significant impact on chalcone yield. The equivalent of NaOH, temperature, and time were also identified as significant factors and selected for DoE analysis. While this one factor at a time (OFAT) approach identified suitable conditions, it did not identify interconnected variables or provide a system-wide analysis. A DoE workflow was established (Fig. 4) to screen multiple variables; each performed in triplicate. This experimental design provided 45 datapoints to determine the optimum conditions for the Claisen–Schmidt condensation reaction¹³ for chalcone synthesis using a system-wide approach.

Fig. 3 Claisen–Schmidt condensation for C1 monitored via LC-MS, starting materials 2-acetylpyridine and benzaldehyde to chalcone C1 with acetaminophen used as an internal standard.

Fig. 4 DoE workflow for optimum chalcone synthesis using 45 individual datapoints.

A least squares fit DoE model was generated with an R² value of 0.97 and a P-value of <0.001, indicating good statistical modelling (Fig. 5C). This model was then used to predict various reaction conditions (Fig. 5A). Solvent was confirmed as influencing the reaction yield, with the polar protic solvents MeOH, EtOH, and IPA providing the best yields. Polar aprotic solvent MeCN and chlorinated solvent dichloromethane should be avoided. Interestingly, the use of 0.2 equivalent NaOH was the most efficient alongside 4 °C reaction temperature. Further increases in catalyst and temperature were detrimental to reaction yield. Time had a minor influence on reaction yield, with 4 h and 8 h the optimal. Using the above model and the 12 principles of green chemistry,³¹ particularly atom economy and energy efficiency, the following conditions were selected: EtOH, 0.2 equiv. NaOH, 4 °C and 8 h. These reaction conditions were then used to synthesise a structurally diverse library of 27 chalcones in excellent yield. Performing these reactions on a 10 mmol scale enabled gram-scale purification of chalcones directly via recrystallization preventing the requirement for time-consuming column chromatography (Scheme 1).

Fig. 5 DoE summary, optimal parameters are displayed in red (panel A), residuals in panel B. An excellent predictive model (R² = 0.97) was generated with confidence interval displayed in red (panel C).

Scheme 1 DoE optimised chalcone synthesis.

The high yield for chalcones C1–C27 provides further confirmation of the predictive properties of the DoE model and validates the chosen reaction conditions (Table 1). These reaction conditions are likely to be suitable for future chalcone research beyond sensing, for example medicinal chemistry.^9–12 These chalcones then served as precursors for a pyrazoline synthesis DoE and, in turn, a pyrazoline library. A suitable DoE approach was developed to determine the optimal reaction conditions for the conversion of chalcone C1 to pyrazoline P1. The same LC-MS method was utilized to monitor the conversion of C1 to P1 with a DoE workflow established to screen solvent, phenylhydrazine equivalent, temperature, and time (Fig. 6).

Fig. 6 DoE workflow for optimum pyrazoline synthesis using 60 individual datapoints.

Table 1 Chalcone library using DoE derived conditions

R1
R2
	C1	C10	C19
	83%	72%	72%
	C2	C11	C20
	72%	73%	76%
	C3	C12	C21
	86%	77%	79%
	C4	C13	C22
	60%	84%	72%
	C5	C14	C23
	75%	60%	70%
	C6	C15	C24
	89%	65%	89%
	C7	C16	C25
	87%	74%	95%
	C8	C17	C26
	92%	84%	76%
	C9	C18	C27
	77%	73%	75%

---

A least squares fit DoE model was generated with EtOH identified as the optimal medium for the reaction (Fig. 7). Interestingly, temperature exerted only a minor influence on pyrazoline formation, with 4 °C providing the best outcome when used in combination with 3.0 equivalents of hydrazine.

Fig. 7 DoE summary, optimal parameters are displayed in red (panel A), residuals in panel B. A good predictive model (R² = 0.93) was generated with confidence interval displayed in red (panel C).

Reaction time showed a linear correlation with product yield, and 24 hours was determined to be optimal. Guided by this optimization model and the 12 Principles of Green Chemistry,³¹ the final reaction conditions were selected as EtOH, 3.0 equiv. hydrazine, 4 °C, and 24 h (Scheme 2). Under these conditions, a structurally diverse pyrazoline library comprising 11 potential sensors was synthesized in good to excellent yields (Fig. 8). Reactions performed on a 2 mmol scale generated sufficient material for purification via recrystallization, eliminating the need for time-consuming column chromatography. This approach was essential for achieving rapid synthesis and enabling high-throughput screening for desirable sensing properties. A preliminary fluorescence screen was performed to identify promising sensors in the presence of Zn²⁺ and Cd²⁺ (Fig. 8). Pyrazoline P1–P3, all containing a phenyl-substituted pyrazoline, displayed negligible changes in fluorescence emission with Zn²⁺ and Cd²⁺. Interestingly, the pyridine-substituted analogues P7–P9 did display a slight preference for Zn²⁺ over Cd²⁺ with green fluorescence emission. This observation suggested that the addition of nitrogen may be beneficial to group 12 metal selectivity. Pyrazoline P11 displayed a “turn off” response, indicating the naphthyl unit is impacting the fluorescence response. The lead sensor was identified as P10, displaying both a nitro-substituted ring and pyridyl units. A highly unusual yellow fluorescence emission was observed in the presence of Zn²⁺, but not with Cd²⁺. The substitution of an electronegative halogen (as observed in P8 and P9) for a highly electronegative nitro group profoundly modified the response to Zn²⁺ and Cd²⁺ and warrants further investigation. With a library of 11 pyrazolines in hand, a high-throughput screening (HTS) assay was developed using a standard laboratory fluorescence plate reader using a 96-well format. The responses to eleven different metal ions (250 μM) across each pyrazoline (50 μM) were measured in duplicate in acetonitrile.

Scheme 2 DoE optimised pyrazoline synthesis.

Fig. 8 Pyrazoline library with initial Zn²⁺ and Cd²⁺ screening using a 100 W 365 nm lamp.

This approach was designed to rapidly screen for fluorescence sensing properties in the 380–650 nm range using 360 nm excitation (see SI for full analysis, see Fig. 9 for heatmap summary). Pyrazoline P1, P3–5, P7–P8, and P10 all displayed a “turn on” response to a range of group 12 metals. The paramagnetic metals Fe³⁺, Ru³⁺, and Cu²⁺ displayed a “turn off” response across all pyrazolines; this is commonly observed in the literature.³² The biological metals Na⁺, Li⁺, and Mg²⁺ did not alter the fluorescence response, and therefore these sensors are unsuitable for monitoring these analytes. The metal screen for P10 is of particular interest as there is a “turn on” response only for Zn²⁺, not Cd²⁺ or Hg²⁺. The Stokes shift for P10 is also unusually large, >240 nm and this is advantageous for sensing applications. This suggests P10 could be a highly useful Zn²⁺ specific sensor and will be explored in future work. In summary, a combined DoE and HTS workflow has been established to rapidly synthesise a pyrazoline library from chalcone precursors in high yield. These sensors were rapidly screened for desirable sensing properties, resulting in the unexpected discovery of sensor P10 with future potential.

Fig. 9 High-throughput screening (HTS) of pyrazoline P1–P11 using a fluorescence plate reader, λ_ex 360 nm, each result is average spectra from duplicate measurements. Colour indicates change in fluorescence, red represents an increase, blue a decrease and green no change in λ_em.

Conclusion

The optimised synthetic conditions for chalcone and pyrazoline formation has been determined using a data-driven design of experiments (DoE) approach (Fig. 10). These conditions (EtOH, 0.2 equiv. NaOH, 4 °C and 8 h) provided a diverse chalcone library of 27 compounds in good to excellent yield (60–93%, Table 1). The optimal conditions for pyrazoline formation (EtOH, 3.0 equiv. hydrazine, 4 °C, and 24 h) were determined and used to generate a pyrazoline library also in good to excellent yield (11–94%, Fig. 8).

Fig. 10 All DoE factors evaluated, with the best conditions in bold.

The establishment of an HTS assay enabled the rapid identification of fluorescent properties to a range of metals, resulting in the highly expected discovery of P10 with highly unusual yellow fluorescent emission in the presence of Zn²⁺ only. This workflow demonstrates the significant advantages of incorporating modern data-driven methodology into synthetic organic chemistry. This approach is ideally suited to machine learning (ML) applications in which a large, structurally diverse, and modular chemical library is synthesized and screened, resulting in large datasets of consistent data.³³ This dataset can then be used to train ML models to predict and design novel compounds with specific properties (Fig. 11A). These, in turn, are generated and screened, and the resulting data used to improve the model further. A key requirement for this strategy is easily accessible scaffolds that are highly modular, using inexpensive reagents. As highlighted within, pyrazolines are ideally suited to this approach. Potential applications of such workflows include the generation of ML-derived sensors for specific metals and the development of molecular logic gates (Fig. 10B).³⁴ Numerous recent pyrazolines have been reported as logic gates,³⁵ and the application of ML will accelerate development further. This is future work and will be reported in due course.

Fig. 11 Future workflow incorporating design of experiments (DoE), high-throughput screen (HTS), and machine learning (ML) to accelerate the discovery process (panel A). Potential outputs include ML-derived sensors for specific metals and the discovery of pyrazoline-based molecular logic gates (panel B).

Author contributions

Alexander Ciupa designed, synthesized, characterised all compounds, and authored the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5ob01684c.

Acknowledgements

Krzysztof Pawlak is acknowledged for assistance with development of the high throughput screening assay and Steven Robinson for assistance with time-of-flight high resolution mass spectrometry. This work made use of shared equipment located at the Materials Innovation Factory, created as part of the UK Research Partnership Innovation Fund (Research England) and co-funded by the Sir Henry Royce Institute.

References

1. Selected examples: (a) A. F. de Almeida, R. Moreira and T. Rodrigues, Nat. Rev. Chem., 2019, 3, 589; (b) L. Himanen, A. Geurts, A. S. Foster and P. Rinke, Adv. Sci., 2019, 6, 1900808; (c) J. Y. Peng, D. Schwalbe-Koda, K. Akkiraju, T. Xie, L. Giordano, Y. Yu, C. J. Eom, J. R. Lunger, D. J. Zheng, R. R. Rao, S. Muy, J. C. Grossman, K. Reuter, R. Gómez-Bombarelli and Y. Shao-Horn, Nat. Rev. Mater., 2022, 7, 991.
2. K. Telford, Johns Hopkins APL Tech. Dig., 2007, 27, 224.
3. A. Ekebergh, C. Lingblom, P. Sandin, C. Wennerås and J. Mårtensson, Org. Biomol. Chem., 2015, 13, 3382.
4. V. Karaluka, R. M. Lanigan, P. M. Murray, M. Badland and T. D. Sheppard, Org. Biomol. Chem., 2015, 13, 10888.
5. P. M. Murray, F. Bellany, L. Benhamou, D.-K. Bučar, A. B. Tabor and T. D. Sheppard, Org. Biomol. Chem., 2016, 14, 2373.
6. J. R. Broach and J. Thorner, Nature, 1996, 384, 14.
7. S. M. Senkan, Nature, 1998, 394, 350.
8. X. Rodríguez-Martínez, E. Pascual-San-José and M. Campoy-Quiles, Energy Environ. Sci., 2021, 14, 3301.
9. P. Singh, A. Anand and V. Kumar, Eur. J. Med. Chem., 2014, 85, 758.
10. Selected examples: (a) C. Karthikeyan, N. S. H. Narayana Moorthy, S. Ramasamy, U. Vanam, E. Manivannan, D. Karunagaran and P. Trivedi, Recent Pat. Anti-Cancer Drug Discovery, 2015, 10, 97; (b) S. Ducki, R. Forrest, J. A. Hadfield, A. Kendall, N. J. Lawrence, A. T. McGown and D. Rennison, Bioorg. Med. Chem. Lett., 1998, 8, 1051; (c) A. Ciupa, N. J. Griffiths, S. K. Light, P. J. Wood and L. Caggiano, Med. Chem. Commun., 2011, 2, 1011.
11. Selected examples: (a) Z. Nowakowska, Eur. J. Med. Chem., 2007, 42, 125; (b) A. Yadav, V. Sharma and G. Singh, ChemistrySelect, 2024, 9, e202401321.
12. D. K. Mahapatra, S. K. Bharti and V. Asati, Eur. J. Med. Chem., 2015, 101, 496.
13. Selected examples: (a) L. Claisen and A. Claparède, Ber. Dtsch. Chem. Ges., 1881, 14, 2460; (b) J. G. Schmidt, Ber. Dtsch. Chem. Ges., 1881, 14, 1459; (c) D. Coutinho, H. G. Machado, V. H. Carvalho-Silva and W. A. da Silva, Phys. Chem. Chem. Phys., 2021, 23, 6738; (d) C. L. Perrin and J. Woo, Phys. Chem. Chem. Phys., 2022, 24, 18978.
14. B. Varghese, S. N. Al-Busafi, F. O. Suliman and S. M. Z. Al-Kindy, RSC Adv., 2017, 7, 46999.
15. D. Matiadis and M. Sagnou, Int. J. Mol. Sci., 2020, 21, 5507.
16. M. Mantzanidou, E. Pontiki and D. Hadjipavlou-Litina, Molecules, 2021, 26, 3439.
17. Selected examples: (a) S. Hofmann, M. Linden, J. Neuner, F. N. Weber and S. R. Waldvogel, Org. Biomol. Chem., 2023, 21, 4694; (b) S. V. Gamapwar, N. P. Tale and N. N. Karade, Synth. Commun., 2012, 42, 2617; (c) A. Ciupa, New J. Chem., 2024, 48, 13900.
18. Selected examples: (a) M. A. Alam, Future Med. Chem., 2023, 15, 2011; (b) T. Sarwar, G. Mustafa, W. Zafar, A. U. Hassan, S. H. Sumrra and A. Asif, J. Mol. Struct., 2025, 1348, 143523.
19. Selected examples: (a) M. Z. Alam, S. Ahmad and S. A. Khan, J. Fluoresc., 2025, 35, 1241; (b) M. Mohasin and S. A. Khan, J. Fluoresc., 2025, 35, 2553; (c) M. Z. Alam, M. Mohasin, S. Ahmad, U. Salma and S. A. Khan, J. Coord. Chem., 2025, 78, 367.
20. Selected examples: (a) Z. Zhang, F. W. Wang, S. Q. Wang, F. Ge, B. X. Zhao and J. Y. Miao, Org. Biomol. Chem., 2012, 10, 8640; (b) R. Manjunath and P. Kannan, New J. Chem., 2018, 42, 10891; (c) Y. S. Yang, F. N. Wang, Y. P. Zhang, F. Yang and J. J. Xue, J. Fluoresc., 2024, 34, 159; (d) Y. P. Zhang, Y. Y. Dong, Y. S. Yang, H. C. Guo, B. X. Cao and S. Q. Sun, Spectrochim. Acta, Part A, 2017, 177, 147; (e) T. T. Zhang, F. W. Wang, M. M. Li, J. T. Liu, J. Y. Miao and B. X. Zhao, Sens. Actuators, B, 2013, 186, 755.
21. Selected examples: (a) S. Gond, P. Yadav, A. Singh, S. Garai, A. Shekher, S. C. Gupta and V. P. Singh, Org. Biomol. Chem., 2023, 21, 4482; (b) A. Tigreros and J. Portilla, Curr. Chin. Sci., 2021, 1, 197; (c) S. Saha, A. De, A. Ghosh, A. Ghosh, K. Bera, K. S. Das, S. Akhtar, N. C. Maiti, A. K. Das, B. B. Das and R. Mondal, RSC Adv., 2021, 11, 10094; (d) A. Dhara, N. Guchhait, I. Mukherjee, A. Mukherjee and S. Chandra Bhattacharya, RSC Adv., 2016, 6, 105930.
22. C. T. Chasapis, C. A. Spiliopoulou, A. C. Loutsidou and M. E. Stefanidou, Arch. Toxicol., 2012, 86, 521.
23. G. Genchi, S. M. Sinicropi, G. Lauria, A. Carocci and A. Catalano, Int. J. Environ. Res. Public Health, 2020, 17, 3782.
24. Y. S. Wu, A. I. Osman, M. Hosny, A. M. Elgarahy, A. S. Eltaweil, D. W. Rooney, Z. Chen, N. S. Rahim, M. Sekar, S. C. B. Gopinath, N. N. Mat Rani, K. Batumalaie and P. S. Yap, ACS Omega, 2024, 9, 5100.
25. Selected examples: (a) Z. Xu, K.-H. Baek, H. N. Kim, J. Cui, X. Qian, D. R. Spring, I. Shin and J. Yoon, J. Am. Chem. Soc., 2010, 132, 601; (b) Y. Tan, J. Gao, J. Yu, Z. Wang, Y. Cui, Y. Yang and G. Qian, Dalton Trans., 2013, 42, 11465; (c) G. Tian, Y.-Z. Han and Q. Yang, J. Mol. Struct., 2023, 1273, 134341.
26. A. Ciupa, M. F. Mahon, P. A. De Bank and L. Caggiano, Org. Biomol. Chem., 2012, 10, 8753.
27. A. Ciupa, RSC Adv., 2024, 14, 3519.
28. A. Ciupa, RSC Adv., 2024, 14, 34918.
29. A. Ciupa, RSC Adv., 2025, 15, 3465.
30. A. Ciupa, Anal. Methods, 2025, 17, 5782.
31. Selected examples: (a) P. Anastas and N. Eghbali, Chem. Soc. Rev., 2010, 39, 301; (b) R. A. Sheldon, Chem. Soc. Rev., 2012, 41, 1437; (c) R. A. Sheldon, Chem. Commun., 2008, 3352.
32. Selected examples: (a) Z. Zhang, F.-W. Wang, S.-Q. Wang, F. Ge, B.-X. Zhao and J.-Y. Miao, Org. Biomol. Chem., 2012, 10, 8640; (b) Z. L. Gong, F. Ge and B.-X. Zhao, Sens. Actuators, B, 2011, 159, 148; (c) M. M. Li, F. Wu, X. Y. Wang, T. T. Zhang, Y. Wu, Y. Xiao, J. Y. Miao and B.-X. Zhao, Anal. Chim. Acta, 2014, 826, 77.
33. J. Vamathevan, D. Clark, P. Czodrowski, I. Dunham, E. Ferran, G. Lee, B. Li, A. Madabhushi, P. Shah, M. Spitzer and S. R. Zhao, Nat. Rev. Drug Discovery, 2019, 18, 463.
34. A. Ciupa, RSC Adv., 2025, 15, 10565.
35. Selected examples: (a) D. C. Magri and A. A. Camilleri, Chem. Commun., 2023, 59, 4459; (b) D. Sammut, N. Bugeja, K. Szaciłowski and D. C. Magri, New J. Chem., 2022, 46, 15042; (c) N. Zerafa, M. Cini and D. C. Magri, Mol. Syst. Des. Eng., 2021, 6, 93.

---