1 Introduction

The creation of the chiral diarylmethanol derivatives has drawn much attention to building blocks in synthesizing medicines, agrochemicals, and other chiral ligands [1, 2]. Moreover, diarylmethanols are significant ingredients and precursors of physiologically active chemicals because of the vital biological activity of their different derivatives. For instance, the diarylmethanol derivative (S)-phenyl(pyridin-2-yl) methanol possesses analgesic and anticonvulsant properties [3]. Neobenodine, orphenadrine, and carbinoxamine have potent antihistaminic effects, as shown in Fig. 1 [4,5,6,7].

Fig. 1
figure 1

Some examples of important aryl heteroaryl chiral compounds

Full size image

In high optical purity, many techniques have been established for building the chiral diarylmethanol core moiety [8, 9]. One of the most effective ways to reduce pro-chiral diaryl ketones is chemical reduction, and different chemical catalysts have been used to achieve extremely enantioselective (> 95% ee) reduction [5, 10, 11]. However, the direct asymmetric hydrogenation of 2-benzoylpyridine is challenging owing to the coordination effect of pyridines [12]. On the other hand, green biocatalytic approaches have become an alternative to chemical techniques in order to produce enantiomerically pure pyridyl phenylmethanols. Also, a promising method is to use biocatalysts, considering the asymmetric reduction of pro-chiral ketones. Some diaryl ketones were reduced with remarkable enantioselectivity by many biological catalysts [13]. However, a limited number of studies in the literature involve the biocatalytic asymmetric reduction of 2-benzoylpyridine [14]. It is also noted that a number of biocatalysts have been applied to the asymmetric reduction of various heterocyclic ketones involving AS2.2241, Candida maris, Candida viswanathii, Daucus carota, Geotrichum candidum, Pseudomonas putida UV4, Rhodococcus ruber DSM 44541 and 4333, Rhodotorula sp., and Rhizopus arrhizus [15,16,17,18,19,20,21,22]. Using the Lactobacillus paracasei BD101 biocatalyst, (S)-2 was reported for the first time in the literature to be obtained in gram scale (5.87 g) with 93% yield as an enantiopure form [23].

The recent studies of obtaining bioreduction conditions are summarized as follows. Particularly, Kavi et al. [24] developed a Box–Behnken design-focused model to find optimum conditions of enantiopure (S)-1-(4-Methoxyphenyl) ethanol with whole cells of Lactobacillus senmaizuke. Further, Aksuoğlu et al. [25] proposed an I-optimal design-embedded model to optimize asymmetric bioreduction optimum conditions of 2-methyl-1-phenylpropan-1-one by Lactobacillus fermentum BY35. Next, Danouche et al. [26] optimized sulfate leaching from Phosphogypsum to find efficient bioreduction. Then, Che et al. [27] used a structure-guided rational design for a ketoreductase from Candida glabrata to acquire the asymmetric reduction of the ketone precursor 2-chloro-1-(3, 4-difluorophenyl) ethanone. Finally, Arnodo et al. [28] reported the enantioselective reduction of 2-substituted cyclic imines associated with amines by imine reductases in non-traditional solvents.

In the literature, the phenyl(pyridin-2-yl)methanone (1) asymmetric reduction was investigated while conducting the classical techniques. As observed from the previous studies, the reaction conditions were obtained where one of them was constant, and other reaction conditions were accordingly found. This technique is not desirable for a couple of reasons. First, this technique did not observe the interaction, quadratic, and higher-order effects of the reaction conditions. Second, obtaining the reaction conditions takes lots of computation time. Third, this method is expensive because many experimental runs may be necessary to acquire the optimum reaction conditions. Based on these awarenesses, a mathematical optimization technique eliminates the disadvantages of the classical approach. The superiority of the mathematical optimization technique is to optimize the reaction conditions simultaneously. In addition, the interaction, quadratic, and higher-order effects are analyzed. Next, the number of experimental runs is less, so the reaction conditions are obtained cheaper and more quickly. Notice that a large amount of substrate is converted into a product based on the usability of the method and biocatalyst, for example, in pharmaceutical industries. The classical method used in the literature may improve the biocatalytic asymmetric reduction of 1 based on yield, substrate amount, and an efficient technique. It is also investigated that a mathematical optimization method has not been conducted to acquire optimum reaction conditions for the biocatalytic reduction of substrate 1.

As observed from the literature, no multi-response nonlinear optimization model is presented to obtain the phenyl(pyridin-2-yl)methanone (1) asymmetric reduction conditions. This paper bridges the research gap between an inscribed design and a nonlinear optimization model for obtaining bioreduction conditions. This paper has three main contributions: First, we selected a biocatalyst, Leuconostoc pseudomesenteroides N13, from Turkish sourdough while catalyzing the enantioselective reduction of 1. Second, the full second-order approximation function was presented to obtain unbiased estimates with restricted maximum likelihood (REML). The estimated responses of ee and conversion were found using the REML analysis. Third, we proposed a novel inscribed design-focused multi-response nonlinear optimization model to obtain optimum bioreduction conditions of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13. In addition, a gradient-based solution technique was applied to solve the proposed nonlinear optimization model. We also tested the usability of the novel inscribed design-focused multi-response nonlinear optimization model, which has not been used in any biocatalytic reactions in the literature for the optimum reaction conditions, as shown in Fig. 2.

Fig. 2
figure 2

Phenyl(pyridin-2-yl)methanone asymmetric reduction using the novel inscribed design-focused multi-response nonlinear optimization model in the presence of Leuconostoc pseudomesenteroides N13 biocatalyst

Full size image

The novel inscribed design-focused multi-response nonlinear optimization model has been successfully used in a biocatalytic asymmetric reduction, and the reaction conditions have been optimized. Under optimized analgesic reaction conditions (S)-2, the conversion, yield, and high gram scale were obtained as 99%, 98%, and 11.79 g, respectively. Along the same lines, this research presents a practical and novel procedure in order to produce (S)-2 with a biocatalyst. In addition, a novel inscribed design-focused multi-response nonlinear optimization model was proposed for the first time in the biocatalytic reduction of substrate 1. (S)-2 was also obtained using the selected biocatalyst for the first time at the most significant gram scale.

The organization of the remainder of this paper is presented as follows: The material and proposed methodology are explained in Sect. 2, including general information, the selected biocatalyst, general and gram-scale procedures, and the proposed optimization methodology. Results and discussions are elaborated on in Sect. 3. To sum up, Sect. 4 is related to the concluding remarks.

2 Material and Proposed Methodology

This section describes the material and proposed methodology and consists of five subsections: (1) General information and material, (2) Biocatalyst (Leuconostoc pseudomesenteroides N13), (3) A general procedure for the production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13, (4) Gram-scale procedure production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13, and (5) A novel inscribed design-focused multi-response nonlinear optimization model for the asymmetric reduction.

2.1 General Information and Material

The materials, such as substrate, solvents, and culture medium (MRS), were purchased by commercial providers. TLC (thin layer chromatography) was employed to track the progress using an ethyl acetate: hexane (05:95) solvent mixture. When using this mixture, (S)-2 was purified using column chromatography. An Agilent 1260 HPLC system was applied to conduct chiral HPLC analysis involving a chiral detector and column. Then, polarimetry (Bellingham + Stanley ADP 220) was used to determine the specific optical rotation of the product. The produced racemic alcohol sample 2 was obtained while reducing 1 with NaBH4 in MeOH. A Bruker (400 MHz) was used in order to record the spectra of the 13C-NMR and 1H-NMR with the solvent CDCl3. Next, (S)-phenyl (pyridin-2-yl)methanol [29]: M. p.: 72–73 °C, White solid, Yield: 98%,1H NMR with 400 MHz and CDCl3, δ = 8.56–8.54 (m, 1H), 7.63–7.59 (m, 1H), 7.40–7.27 (m, 5H), 7.20–7.15 (m, 1H), 5.76 (s, 1H), 5.38 (bs, 1H (OH)); 13C NMR with 100 MHz and CDCl3, δ = 160.9, 147.8, 143.2, 136.8, 128.6, 127.8, 127.1, 122.4, 121.3, 75.0; [α]D25 = 129.3 for c 1.684 and CHCl3 > 99% ee; Chiralcel AD column, HPLC, hexane/i‐PrOH, 90:10, 215 nm, flow rate = 0.8 mL/min, tR (S) = 15.8 min. As provided in the supporting information, the HPLC condition of 1 is denoted as follows: Chiralcel AD column, 90:10, the flow rate of 0.8 mL/min, n-hexane/i-PrOH, 215 nm, and 9.3 min.

2.2 Biocatalyst (Leuconostoc pseudomesenteroides N13)

Leuconostoc pseudomesenteroides N13 was previously identified and preserved in the lab from the Turkish sourdough. As previously described in the literature, the biocatalyst was isolated, dried, and stored [30].

2.3 A General Procedure for the Production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13

Dry Leuconostoc pseudomesenteroides N13 (25 mg) was added to Erlen Mayer (250 mL) involving MRS broth (100 mL), which was freshly prepared, and the mixture was agitated at 153 rpm for two hours at 29 °C. The reaction medium's pH was 1 M HCl to 6.0. And then, the mixture was stirred for an extra two hours. After, the mixture was added to 1 mmol of 1 and went for 53 h at 29 °C and 153 rpm. Then, the mixture was centrifuged, and the liquid part was separated. The liquid phase was extracted in order to determine the conversion rate, and then, the organic layer was dried and dissolved with the isopropanol. After a small amount of crude product was filtered through a small silica column, the sample was used to detect the conversion rate and the ee values based on the chiral HPLC analysis. Next, the crude product was purified in column chromatography with ethyl acetate: hexane solvent mixture (05:95). The NMR analysis was performed to characterize the product. Comparing the optical rotation values to the literature value allowed us to ascertain their absolute configuration. The reaction transformation was examined by comparing the ketone peak with the alcohol peaks. The product could not be obtained in the trials without the use of a biocatalyst. Notice that the relevant spectra can be seen in the supporting information.

2.4 Gram-Scale Procedure Production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13

Dry Leuconostoc pseudomesenteroides N13 (250 mg) was added in 5-L Erlen Mayer containing 1000 mL MRS broth, which was freshly prepared, and the mixture was agitated (153 rpm) for two hours at 29 °C. Then, the reaction medium's pH was 1 M HCl to 6.0, and the mixture was stirred for an additional two hours. After that, the mixture was added to 65 mmol of 1 and stirred for 53 h at 29 °C and 153 rpm. As a result of the reaction, conversion, ee, and purification of the product were carried out as stated above. Characterization of the product, conversion of the substrate, and ee of the product were carried out as described above.

2.5 A Novel Inscribed Design-Focused Multi-response Nonlinear Optimization Model for the Asymmetric Reduction

The details of each step are explained as follows. The first step is planning, and the experimenters decide the design-related issues, including which variables are hard or easy to change. The levels of one or more design variables may be hard to change for an experiment. Thus, the design should be split into a whole plot and subplot. Further, the response surface design is appropriate for building a second-order approximation response [31,32,33]. The second-order response function has three main advantages as follows [31]: (1) flexibility, (2) easy estimations, and (3) practicality for actual experiments. Therefore, a novel inscribed design is performed when combining the split plot and response surface approaches. In addition, the presented design is rotatable in this paper to provide prediction variance without changing the experimental design space. Also, the inscribed design provided the process stability as another property. Thus, these properties are important in choosing the inscribed design in this paper over the other experimental methods.

The design variables and two responses are denoted in Table 1 for the experiment. The design variables and their coded and uncoded levels were specified based on the experimenters' knowledge, experience, and expectations. Also, the specified coded and uncoded levels are consistent with the previous studies. Note that adjusting pH may be complicated for this particular experiment. Therefore, the pH is defined as the hard-to-change design variable. An inscribed design consists of factorial, center, and axial points. Due to the nature of the inscribed design, eight factorial points were used in each group to observe the main and interaction effects. Next, three center points were enough to run for each center point group when providing process stability. In addition, the total number of axial points was specified as ten in Table 2, and they were sufficient to allow curvature estimation. Therefore, as shown in Table 2, thirty-five experimental runs were conducted, including three groups of center points, two groups of factorial points, two groups of the hard-to-change design variable of axial points, and one group of the three easy-to-change design variables of axial points. Each experimental run was randomly performed in order to eliminate the experimental prejudices.

Table 1 The four design variables and two responses
Full size table
Table 2 The thirty-five runs inscribed design for the experiment
Full size table

The finding of the estimated response functions is another important step in analyzing significant model parameters. For this particular purpose, the REML analysis was conducted for each response. Also, unbiased estimates are provided by using the REML. In this paper, the full second-order approximation function is used to estimate the model terms for the response. Therefore, the estimated fitted response functions, \(\hat{\mu }_{1}\) and \(\hat{\mu }_{2}\), are denoted for the ee and conversion, respectively, as follows:

$$\begin{aligned} \hat{\mu }_{1} = & \hat{\beta }_{0} + \hat{\beta }_{1} E + \hat{\beta }_{2} F + \hat{\beta }_{3} G + \hat{\beta }_{4} H \\ &+ \hat{\beta }_{12} EF + \hat{\beta }_{13} EG + \hat{\beta }_{14} EH + \hat{\beta }_{23} FG \\ \quad & + \hat{\beta }_{24} FH + \hat{\beta }_{34} GH + \hat{\beta }_{11} E^{2} \\ &+ \hat{\beta }_{22} F^{2} + \hat{\beta }_{33} G^{2} + \hat{\beta }_{44} H^{2} + \varepsilon_{1} \\ \end{aligned}$$
(1)
$$\begin{aligned} \hat{\mu }_{2} =& \hat{\delta }_{0} + \hat{\delta }_{1} E + \hat{\delta }_{2} F + \hat{\delta }_{3} G + \hat{\delta }_{4} H + \hat{\delta }_{12} EF \\ &+ \hat{\delta }_{13} EG + \hat{\delta }_{14} EH + \hat{\delta }_{23} FG \hfill \\& + \hat{\delta }_{24} FH + \hat{\delta }_{34} GH + \hat{\delta }_{11} E^{2} \\ & + \hat{\delta }_{22} F^{2} + \hat{\delta }_{33} G^{2} + \hat{\delta }_{44} H^{2} + \varepsilon_{2} \hfill \\ \end{aligned}$$
(2)

As indicated in (1) and (2), two response functions are obtained from the inscribed design. The objective of the novel inscribed multi-response nonlinear optimization model is to maximize the two response functions; therefore, objective functions are defined in terms of the desirability of each response. In addition, the design space limitations of the four design variables should be incorporated in order to prevent meaningless results. Further, each desirability value should be defined between zero and one. The highest value of desirability is wanted for each response. The novel inscribed design-focused multi-response nonlinear optimization model is denoted in order to obtain optimum bioreduction conditions of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13 as follows:

$$\begin{array}{*{20}l} {{\text{maximize }}\left\{ {d\left( {\hat{\mu }_{1} } \right),d\left( {\hat{\mu }_{2} } \right)} \right\}} \hfill \\ {{\text{subject to }} - 1 \le E,F,G,H \le + 1} \hfill \\ {\text{where }} \hfill \\ {d\left( {\hat{\mu }_{1} } \right) = \left\{ {\begin{array}{*{20}l} {0, \, \hat{\mu }_{1} {\text{ < lower}}_{{{\text{ee}}}} } \hfill \\ {\left( {\frac{{\hat{\mu }_{1} - {\text{lower}}_{{{\text{ee}}}} }}{{{\text{target}}_{{{\text{ee}}}} - {\text{lower}}_{{{\text{ee}}}} }}} \right),{\text{ lower}}_{{{\text{ee}}}} \le \hat{\mu }_{1} \le {\text{target}}_{{{\text{ee}}}} } \hfill \\ {1, \, \hat{\mu }_{1} > {\text{lower}}_{{{\text{ee}}}} } \hfill \\ \end{array} } \right..} \hfill \\ {d\left( {\hat{\mu }_{2} } \right) = \left\{ {\begin{array}{*{20}l} {0, \, \hat{\mu }_{2} {\text{ < lower}}_{{\text{c}}} } \hfill \\ {\left( {\frac{{\hat{\mu }_{2} - {\text{lower}}_{{\text{c}}} }}{{{\text{target}}_{{\text{c}}} - {\text{lower}}_{{\text{c}}} }}} \right),{\text{ lower}}_{{\text{c}}} \le \hat{\mu }_{2} \le {\text{target}}_{{\text{c}}} } \hfill \\ {1, \, \hat{\mu }_{2} > {\text{lower}}_{{\text{c}}} } \hfill \\ \end{array} } \right.,} \hfill \\ {{\text{Solution }}\;{\text{technique: A }}\;{\text{gradient - based}}\;{\text{ procedure}}} \hfill \\ {{\text{Find: Optimum }}\;{\text{bioreduction }}\;{\text{conditions}}} \hfill \\ \end{array}$$
(3)

Design Expert 12.0.3.0 (Minneapolis, MN 55143) was applied to statistical analysis and plotting the graphs. Next, a gradient-based solution procedure was used to solve the novel inscribed design-focused multi-response nonlinear optimization model in (3). For this particular purpose, MATLAB R2014a (Natick, Massachusetts 01760) with an optimization toolbox was run in order to solve the proposed optimization model.

3 Results and Discussion

This section consists of three subsections as follows: (1) the results and discussion of the statistical analysis, (2) the results and discussion of the novel inscribed design-focused multi-response nonlinear optimization model, and (3) Gram-scale production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13.

3.1 The Results and Discussion of the Statistical Analysis

The statistical analysis was conducted for the ee. R2 (the maximum coefficient of determination) and \(R_{{{\text{adj}}}}^{2}\) (the adjusted coefficient of determination) values are 0.9837 and 0.9693, respectively. Notice that R2 and \(R_{{{\text{adj}}}}^{2}\) values are 0.9385 and 0.9254, respectively, when dealing with the linear regression model. Based on these values, the quadratic (second-order) regression model is better than the linear regression model. Also, a good model fit is provided associated with the R2 and \(R_{{{\text{adj}}}}^{2}\) values of the quadratic regression model for the ee response. The REML (restricted maximum likelihood) analysis and variance components are depicted in Table 3. Based on the whole plot and subplot, the model is significant. In this model, E (pH), F (temperature), H (agitation speed), and F2 (temperature) are significant model terms based on the Kenward–Roger p-values, where p-values are less than 0.05. Next, the group variance component accounts for 90.35% of the total variability, as shown in Table 3.

Table 3 REML analysis and variance components for the ee response
Full size table

The full second-order approximation response function of the ee is obtained as follows:

$$\begin{aligned} \hat{\mu }_{1} = & 84.00 + 7.73E - 1.46F - 0.13G + 0.71H \\ & - 0.31EF + 0.81EG - 0.06EH \\ &+ 0.44FG + 0.31FH - 0.31GH \\ &- 0.41E^{2} - 3.77F^{2} - 1.64G^{2} - 1.77H^{2} \\ \end{aligned}$$
(4)

The conversion response was statistically analyzed. Particularly, R2 and \(R_{{{\text{adj}}}}^{2}\) values are 0.9655 and 0.9347, respectively. Note that R2 and \(R_{{{\text{adj}}}}^{2}\) values are 0.8655 and 0.8366, respectively, when the regression model is linear. In terms of R2 and \(R_{{{\text{adj}}}}^{2}\) values, the quadratic (second-order) regression model is superior to the linear regression model. In addition, a good model fit is obtained based on the R2 and \(R_{{{\text{adj}}}}^{2}\) values of the quadratic regression model for the conversion response. In Table 4, the REML analysis and variance components are denoted. The model is significant based on the REML analysis, including the whole part and subplot. In the model, E (pH), F (temperature), G (incubation period), H (agitation speed), EH (Ph* agitation speed), E2 (Ph2), F2 (temperature2), G2 (incubation period2), and H2 (agitation speed2) are significant model terms associated with the Kenward–Roger p-values, where the p-values are specified as less than 0.05. Then, the group variance component explains 55.48% of the total variability, as depicted in Table 3.

Table 4 REML analysis and variance components for the conversion response
Full size table

The full second-order approximation response function of the conversion is obtained as follows:

$$\begin{aligned} \hat{\mu }_{2} &= 87.00 + 9.07E - 2.42F + 3.67G + 4.50H - 1.50EF \\ & \quad - 0.75EG + 1.75EH + 0.63FG - 0.13FH \\ & \quad - 0.63GH - 2.11E^{2} - 5.91F^{2}- 2.29G^{2} - 2.54H^{2} \\ \end{aligned}$$
(5)

3.2 The Results and Discussion of the Novel Inscribed Design-Focused Multi-response Nonlinear Optimization Model

The novel inscribed multi-response nonlinear optimization model was effectively applied in finding the optimum bioreduction conditions of (S)-phenyl (pyridin-2-yl)methanol using Leuconostoc pseudomesenteroides N13. Based on the optimization results, each desirability function was obtained one for the ee and conversion. It was reported that the most desirable responses were acquired for the experiment. The bioreduction conditions were acquired as follows: E (pH) = 6.0, F (temperature) = 29 °C, G (incubation period) = 53 h, and H (agitation speed) = 153 rpm. In addition, the estimated values of the ee and conversion were calculated as 98.4% and 99%, respectively.

The validation of the results is necessary to verify bioreduction conditions from the proposed multi-response nonlinear optimization model. Thus, an experiment called validation was performed using the optimum bioreduction conditions of the four design variables from the novel inscribed multi-response nonlinear optimization model. The results of the validation experiment showed that the ee and conversion were investigated as > 99% and > %99, respectively. It is examined that the validation experiment results are consistent with the novel inscribed multi-response nonlinear optimization model results when analyzing the ee, conversion, and yield. It is also concluded that the novel inscribed multi-response nonlinear optimization model is valuable and practical for the asymmetric bioreduction of (S)-phenyl (pyridin-2-yl)methanol using Leuconostoc pseudomesenteroides N13.

Figures 3, 4, 5 show the contour plots of the desirability, ee, and conversion while dealing with the paired design variables. As depicted in Fig. 1, a high pH level is desirable in order to acquire the highest desirability. On the other hand, a low-temperature level is desirable for asymmetric reduction. Further, the incubation period is around 53 h in order to obtain the most desirable for the experiment. Also, the agitation speed is desirable around the medium level, as observed in Fig. 1. It is investigated that the maximized ee and conversion values are obtained at the high pH level, a low-temperature level, an upper middle-temperature level, and around a medium agitation speed level, as well in Figs. 2 and 3. Therefore, desirability is efficiently acquired in order to maximize the estimated values of the ee and conversion. Notice that observations from Figs. 3, 4, 5 are in agreement with the novel inscribed multi-response nonlinear optimization results.

Fig. 3
figure 3

Contour plots of the desirability

Full size image
Fig. 4
figure 4

Contour plots of the ee response

Full size image
Fig. 5
figure 5

Contour plots of the conversion

Full size image

3.3 Gram-Scale Production of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13

The industrial applicability of the method and biocatalyst used in biocatalytic asymmetric reduction studies is significant. Therefore, to demonstrate the industrial applicability of the technique and the biocatalyst, a large amount of the substrate must be converted into the product under optimized conditions. Note that substrate concentration affects selectivity, conversion, and ee [15]. Therefore, various experiments were carried out using various amounts of the substrate under optimized bioreduction conditions in order to convert the maximum amount of substrate into the product. When using 65 mmol (11.9 g) of the substrate under optimized conditions, (S)-2 was obtained as 11.79 g with > 99% conversion and 98% yield as enantiopure form, as depicted in Fig. 6.

Fig. 6
figure 6

Gram-scale production of (S)-2

Full size image

When the amount of substrate was more than 65 mmol, both the selectivity and conversion decreased. For example, when the amount of substrate was 75 mmol, the reaction conversion was 81%, and the ee was 89% for the product. The reason is the substrate toxicity, and biocatalyst decreased activity, as presented in the literature [30].

4 Conclusion

In this study, the enantioselective reduction of aryl heteroaryl ketone 1 catalyzed by Leuconostoc pseudomesenteroides N13 was performed effectively with a novel inscribed design-focused multi-response nonlinear optimization model to give the corresponding enantiomerically pure alcohol (S)-2. Note that the proposed multi-response nonlinear optimization model has not been used in the literature to obtain optimum reaction conditions when dealing with the REML analysis for estimating the ee and conversion responses. Also, it has been shown that the novel inscribed design-focused multi-response nonlinear optimization model can be successfully applied in biocatalytic reactions. After the optimization phase, the bioreduction conditions of (S)-phenyl (pyridin-2-yl)methanol by Leuconostoc pseudomesenteroides N13 were found as: pH = 6.0, temperature = 29 °C, incubation period = 53 h, and agitation speed = 153 rpm. After the validation study, the ee and conversion were obtained as > 99% and > %99, respectively. Further, (S)-2 was obtained as 11.79 g with > 99% conversion and 98% yield as enantiopure form when 65 mmol (11.9 g) of the substrate was used under optimum bioreduction conditions. In addition, this research indicates that Leuconostoc pseudomesenteroides N13 is an effective aryl heteraryl ketone reduction biocatalyst. It is still a difficult task in organic synthesis to reduce diaryl ketones in a high ee, but this biocatalyst and the novel inscribed design-focused multi-response nonlinear optimization model can hold great promise in this regard. Next, a green and efficient biocatalytic strategy was presented in order to produce a pharmaceutically important analgesic, enantiopure (S)-2, which was firstly developed by the inscribed design-focused multi-response nonlinear optimization model with the excellent conversion, ee, yield, and high gram scale. Finally, this procedure has economic potential for the practical synthesis of pharmaceutically effective compounds since it efficiently produces (S)-2.

The limitation of this study is that only the controllable design variables are considered. Besides, the experimental procedure in this paper did not consider noise design variables. On the other hand, both controllable and noise design variables could be considered for asymmetric bioreductions. Another possible future work is developing a nonlinear optimization model when considering both controllable and noise design variables for asymmetric bioreductions. Further, the biocatalyst in this paper could be carried out with different biocatalysts for future research.

Supplementary Data

1H-NMR and 13C-NMR spectrums and chiral and racemic HPLC chromatograms of the product (S)-2 could be found in the supplementary data.