Green extraction optimization of antioxidant bioactives from Spondias tuberosa agroindustrial waste

1 Introduction

Spondias tuberosa (umbu) is an endemic fruit of the Brazilian Caatinga biome. Its commercialization is primarily directed toward fresh consumption and the production of pulp, juices, jams, and other desserts (Oliveira et al., 2024; Gualberto et al., 2021). Its popularity, driven by its sweet flavor with slight acidity, makes it attractive to both the national and international market (Oliveira et al., 2024; Gualberto et al., 2021). The residue generated during fruit processing corresponds to 20%–60% of the raw material. These residues, composed of peel, seeds, and pulp remnants, are rich in vitamin C and secondary metabolites such as flavonoids, tannins, and anthocyanins, which confer antioxidant and anti-inflammatory activities and represent relevant functional properties for food, pharmaceutical, and cosmetic applications (Gualberto et al., 2021; Ribeiro et al., 2022).

Polyphenols are present in plant matrices in both soluble forms, which are freely available, and insoluble forms, bound to structural components of the cell walls, making their recovery more challenging (Zaghbib et al., 2026). Traditional extraction methods involve large sample volumes and the use of volatile, toxic, and flammable organic solvents, generating substantial amounts of environmentally unfriendly waste. In addition, these methods often require complex and time-consuming protocols and are associated with high energy consumption. Green extraction involves the search for and development of methodologies that employ more environmentally friendly solvents, promoting the production of renewable natural products with improved quality (Zaghbib et al., 2026; Palos-Hernández et al., 2024; Žlabur et al., 2020).

Given the imperative to upcycle this agro-industrial by-product within a circular economy framework, the extraction of its functional compounds must be carefully designed (dos Santos et al., 2025). In this context, the extraction solvent plays a critical role due to its affinity for bioactive compounds. The search for eco-friendly solvents and the study of hydroalcoholic systems in different proportions become essential not only to maximize the solubilization of phenolic compounds but also to ensure a green extraction process (Plaskova and Mlcek, 2023). Hydroalcoholic solvents, particularly ethanol–water mixtures, have been widely employed for the extraction of bioactive compounds from plant matrices due to their high efficiency in solubilizing phenolic metabolites, low toxicity, reduced cost, and compatibility with food and pharmaceutical applications (Gualberto et al., 2021; Plaskova and Mlcek, 2023).

Due to the multifactorial nature of extraction processes, univariate approaches are limited in evaluating interactions between variables and defining optimal conditions. Therefore, multivariate statistical tools, such as full factorial design and Response Surface Methodology (RSM), have been widely applied to optimize extraction processes of natural products. These strategies allow the evaluation of individual effects and factor interactions, reduce the number of experiments, and mathematically model the responses, enabling the determination of optimal conditions with greater precision. RSM has been demonstrated to be an effective, robust, and highly reproducible tool for the improvement and optimization of existing extraction methodologies (Mansour et al., 2023).

Driven by the rising generation of waste from the frozen pulp industry and the growing demand for natural antioxidant compounds, this study aimed to optimize the extraction process of bioactives from umbu residues using a mixture of ethanol-water as a green solvent.

2 Materials and methods

2.1 Obtaining Spondias tuberosa residue

The residue, composed of peel and pulp from umbu, was obtained in Manoel Vitorino (Latitude: 14° 08′42″S and Longitude: 40° 14′34″W), Bahia, Brazil. Access to the genetic heritage was registered in SisGen (A96F85C).

2.2 Preparation of Spondias tuberosa residue extract

The residues were dried in a forced-air circulation oven (SolidSteel, SSD 180L) (50 °C/72 h) and later ground using a knife mill and sieved through a 180 µm mesh to standardize particle size. Homogenizer-assisted extraction (turbolysis) was carried out using a domestic blender (KD Eletro, LAR-2), with a power output of 800 W and a speed of 18,000 RPM, and a two-liter capacity. Ultrasound-assisted extraction was performed using a ultrasonic cleaner (Sanders do Brasil, Soniclean 2) operating at a frequency of 40 kHz. The samples, consisting of residue and solvent, were placed in beakers covered with Parafilm and positioned at the center of the ultrasonic bath. The water level in the equipment was sufficient to keep the samples submerged throughout the extraction process. Both extraction procedures were conducted at room temperature (25 °C) and with a residue-to-solvent ratio of 1:20 (w/v) calculated based on the raw matrix for both fresh and dried states. The extraction procedure was carried out according to the experimental matrix defined by the design of experiments. After extraction, the samples were centrifuged (Nova Técnica, NT 809) at 3,000 rpm for 10 min. The supernatant was then vacuum-filtered and stored for subsequent analyses. No experimental replicates were performed for the full factorial design, whereas replicates were included only at the center point of the central composite design.

2.3 Preliminary evaluation of the Spondias tuberosa residue extraction process

A 2⁴ Full Factorial Design (Table 1) was applied to evaluate the extraction process. The independent variables were residue type, extraction method, extraction time and solvent. A total of sixteen runs were conducted in a single block. Extracts obtained with 50% ethanol were concentrated using a rotary evaporator at 45 °C and subsequently subjected to response-variable analyses: DPPH assay, Total Flavonoid Content (TFC), and Total Solid Content (TS), all of which performed in triplicate.

TABLE 1. Full factorial design 2⁴ matrix and results for the extraction of Spondias tuberosa residue.

HAE, Homogenizer-assisted extraction; UAE, ultrasound-assisted extraction; EtOH, ethanol. n.d.: not detected; *µg QE/mg.

2.4 Optimization of the extraction process of Spondias tuberosa residue

For the optimization stage, a Central Composite Design 2² + 2 axial points were performed (Table 2). The independent variables were the extraction solvent (with different ethanol concentrations) and extraction time. A total of eleven experimental runs were conducted in a single block. The experimental design comprised nine unique treatment combinations, including three replicates at the center point to evaluate the pure error. The response variables remained the same.

TABLE 2. Optimization phase of the Spondias tuberosa residue extraction process applying a central composite design 2² + 2 and axial points.

2.5 Characterization of the obtained extracts

2.5.1 Total solid content

For this assay, 3 mL of the plant extract were dried in a forced-air circulation oven at 100 °C for 3 h. The dry residue mass was determined by the difference between the final mass of the set and the initial mass of the empty container. TS content was expressed in mg/mL as the ratio between the dry residue mass and the initial extract volume (3.0 mL) (Ag ência Nacional de Vigilância Sanitária ANVISA, 2023).

2.5.2 DPPH (2,2-diphenyl-1-picrylhydrazyl) radical scavenging assay

For the analysis, 200 µL of ethanolic DPPH solution (0.1 mM) and 100 µL of the extract samples. To avoid concentration bias, all extracts were standardized to the lowest dry residue concentration obtained in each experimental phase: 6.20 mg/mL for the preliminary phase and 9.82 mg/mL for the optimization phase. The samples and the reagent were vortex-mixed for 60 s. Subsequently, the reaction mixtures were incubated in the dark for 30 min. The blank solution consisted of 3 mL of ethanol and 1 mL of water, whereas the negative control consisted of 3 mL of the ethanolic DPPH solution and 1 mL of water. Absorbance was measured at 517 nm, and results were expressed as mean ± standard deviation based on the percentage of radical scavenging. The DPPH radical scavenging assay was calculated according with Equation 1 (Brand-Williams et al., 1995).

D P P H R S A % = 1 − A b s s a m p l e A b s n e g a t i v e c o n t r o l × 100 (1)

2.5.3 Total flavonoid content

Total flavonoid content (TFC) was determined by the aluminum chloride (AlCl₃) complexation method, using quercetin as the standard. For the analysis, 1 mL of 2.5% AlCl₃ solution and 1 mL of extract sample. All extracts were standardized according to 3.5.2. The TFC was calculated using the calibration curve of quercetin (5–35 μg/mL |y = 0,0366x - 0,1063 |R² = 0,9946). The blank solution consisted of 1 mL of methanol and 1 mL of water. Absorbance was measured at 425 nm using a spectrophotometer, and results were calculated and expressed as mean ± standard deviation (µg QE/mg) (Schmidt and Ortega, 1983).

2.5.4 Statistical analysis

The statistical factorial design was developed using the Statistica 12 software (StatSoft Inc., USA). The statistical analysis was performed in the same software using Analysis of Variance (ANOVA) and the Pareto chart. Response surface plots were also generated to evaluate the effects of the independent variables on the dependent responses. The extraction process optimization followed a design with triplicates at the center points, all other analytical characterizations (TS, TFC e DPPH assay) were conducted in independent triplicate.

3 Results

3.1 Preliminary evaluation of the Spondias tuberosa residue extraction process

The TS of umbu residues in the preliminary stage ranged from 0–23.8 mg/mL (Table 1). Although dried residue provided higher extraction efficiency, the analysis of variance was not statistically significant. Similarly, the extraction method did not show statistically significant variance. However, statistically significant effects were observed for extraction time (p = 0.023) and solvent type (p = 0.000018). According to the Response Surface Methodology (RSM) (Figure 1), longer extraction time and the use of water as the solvent contributed to higher extraction efficiency.

Regarding TFC, values ranged from 1.10–23.50 µg QE/mg. The individual factors of extraction time and solvent type showed statistically significant effects. For extraction time, a p = 0.0073 was observed, while solvent type presented a p = 0.00023. Thus, an extraction time of 15 min and the use of water as solvent favored flavonoid recovery (Figure 1). Additionally, the interaction between extraction time and solvent type was significant (p = 0.005418), indicating a synergistic effect between these factors. This synergistic effect altered the individual trends, demonstrating that the maximum flavonoid content was achieved when a 15-min extraction was combined with 50% ethanol (Figure 1).

In the DPPH assay, antioxidant capacity ranged from 14%–90%. Unlike the other dependent variables, this assay did not show statistically significant variance for any of the evaluated factors.

Following a rigorous model refinement procedure to eliminate non-significant effects, the adjusted coefficients of determination (R²adj) stabilized at 89.443% for TFC and 80.219% for TS, with ordinary R² values of 91.554% and 82.857%, respectively. The minimal gap between these ordinary and adjusted coefficients (<2.64%) mathematically demonstrates that the experimental noise was successfully eliminated, validating the descriptive robustness of the refined screening frameworks. The adequacy of these fitted preliminary models was visually confirmed through diagnostic residual analyses included in the Supplementary Material.

3.2 Optimization of the extraction process of Spondias tuberosa residues

During the optimization stage, the extraction method (HAE) and residue type (dried) were fixed, while solvent concentration and extraction time were selected as critical variables. TFC ranged from 8.76 to 29.76 µg QE/mg (Table 2), with statistically significant variation regarding extraction time (p = 0.006) and the linear effect of solvent concentration (p = 0.007), both of which show a positive effect. Therefore, longer extraction times and higher ethanol concentrations favored flavonoid recovery (Figure 2).

TS ranged from 9.82 to 32.71 mg/mL (Table 2). As illustrated by the Pareto chart (Figure 2), the linear and quadratic effects of extraction time, as well as the quadratic effect of solvent concentration, were statistically significant, with p-values of 0.0029, 0.0034, and 0.0033, respectively.

Analysis of variance (ANOVA) revealed a statistically significant lack of fit for both the TS and TFC models (p < 0.05). It is likely that the second-order polynomial model does not fully capture the complexity associated with the heterogeneous nature of the umbu residue matrix, which may explain the observed residual variability. Nevertheless, despite this statistical deviation, the models identified significant main effects of the independent variables, providing a useful approximation of the trends observed throughout the extraction process.

The DPPH assay showed a significant quadratic effect of extraction time (p = 0.026), with values ranging from 84.2% to 95.59%, resulting in a parabolic response surface profile (Figure 2). After executing the model refinement procedure for the optimization phase, the adjusted coefficients of determination (R²adj) were consolidated at 39.151% for TFC, 45.655% for TS, and 46.637% for DPPH, with corresponding ordinary R ² values of 51.320%, 61.958%, and 51.974%, respectively.

The refined model for DPPH scavenging activity exhibited a completely non-significant lack of fit (p = 0.203,731), mathematically validating its capacity to map the antioxidant bioactivity curvature with high predictive fidelity. The underlying assumptions for this functional response were validated by diagnostic plots, showing a normal distribution of errors and a homoscedastic residual scatter profile (Supplementary Material).

Conversely, a statistically significant lack of fit remained mathematically present for both the TFC and TS models. As discussed in the subsequent section, this behavior represents a well-known statistical artifact driven by an exceptionally low variance among the experimental replicates at the center point (near-zero Pure Error), rather than a failure of the surface trends to describe the extraction behavior. This variability is likely associated with the inherent complexity of the plant matrix and the heterogeneity of bioactive compounds within the umbu residue. These values are considered acceptable given the high complexity of plant matrices, the intrinsic variability of secondary metabolites, and their interactions with the cell wall. Overall, the central points of solvent concentration, with 50% ethanol and an extraction time of 10 min, showed better results.

4 Discussion

4.1 Preliminary evaluation of the Spondias tuberosa residue extraction process

During the preliminary stage of the residue extraction process, the extracts obtained from dried material showed higher recovery of solids and flavonoids, as well as enhanced antioxidant activity as evaluated by the DPPH assay. This behavior may be associated with the disruption of cellular matrix structures during the drying process. Milling of the plant material after drying increases the surface area and promotes greater contact between the plant matrix and the extraction solvent, resulting in higher concentrations of bioactive compounds (Plaskova and Mlcek, 2023). This structural disruption facilitates solvent access to secondary metabolites. Additionally, drying promotes the denaturation of oxidative enzymes (e.g., lipoxygenase), which are abundant in plant tissues and are responsible for polyphenol degradation (Plaskova and Mlcek, 2023; Yi and Wetzstein, 2011).

The negligible or near-zero extraction yields observed in extracts obtained from fresh residues can be attributed to the presence of water, which hinders the penetration of the extraction solvent into the cellular matrix. Consequently, the solubilization of less polar antioxidant metabolites, such as flavanones, flavonols (aglycones), and isoflavones, is reduced. Furthermore, oxidative enzymes remain active in the fresh matrix and may contribute to the degradation of extractable compounds. These “n.d.” results do not represent the actual mass of the extracts but rather result from the measurement uncertainty of the analytical balance. For the statistical treatment of the experimental data, these values were considered as “zero”. The combined effect of these factors results in lower solid extraction efficiency, with the resulting yields often approaching the detection limit of the analytical balance (Plaskova and Mlcek, 2023; Yi and Wetzstein, 2011).

Yi and Wetzstein (Yi and Wetzstein, 2011) reported higher polyphenol recovery from dried Mentha piperita leaves extracted with 80% ethanol, likely due to ethanol’s selectivity for less polar flavonoids. In contrast, Bulgaru et al. (Bulgaru et al., 2024) observed a reduced flavonoid content in dried Ribes × nidigrolaria, decreasing from 3.61 to 1.64 mg RuE/g DW with a 60% hydroalcoholic solvent. Antioxidant activity also declined from 44.16 to 20.91 mg TE/g DW under the same conditions, highlighting that drying effects vary according to the plant matrix and extraction conditions.

It is important to note that the initial extraction ratio (1:20 w/v) was applied to the matrices as received. Although the moisture content was not systematically quantified, representing a limitation in fully decoupling moisture dilution from absolute extraction kinetics, all subsequent analytical dosing was normalized on a dry-extract basis to ensure comparative validity.

Although the extraction method was not statistically significant during the preliminary stage, HAE showed a numerical tendency toward enhanced performance, achieving the highest TS (23.8 mg/mL, extract S3) and satisfactory flavonoid extraction efficiency (18.31 µg QE/mg, extract S3). However, the highest TFC (23.5 µg QE/mg) was obtained by UAE (extract S12), with antioxidant activity reaching 90%. High blade rotation speeds generate intense shear forces, whose mechanical energy disrupts the plant matrix and reduces particle size, enabling improved solvent flow and an increased contact surface area, thereby enhancing extraction efficiency (Li et al., 2021).

Li et al. (2021) compared multiple extraction methods for Morus alba L. and observed the superiority of high-speed shear homogenization extraction with hot water (HSEHE) compared to ultrasound-assisted hot water extraction (UAHE). The extract obtained by HSEHE showed an antioxidant activity of 93.67% in the DPPH assay, whereas UAHE showed 80% at a concentration of 1 mg/mL. For yield evaluation, the authors analyzed polysaccharide extraction using different particle sizes (60, 80, and 100 mesh), obtaining yields of 9.95% for HSEHE and 3.64% for UAHE at 100 mesh.

Viell et al. (2020) demonstrated higher flavonoid recovery using UAE due to cavitation-induced shear forces that enhance solvent penetration and diffusion. Using a previously optimized ternary solvent system (49:26:25 water/ethanol/methanol, v/v), the extract of Eragrostis tef reached 106.03 mg CE·100/g, while HAE yielded 77.64 mg CE·100/g. However, for the umbu residue, HAE proved more appropriate, as the intense shear stress disrupts its complex fibrous matrix, reduces particle size, and increases the accessible surface area. This mechanical effect improves mass transfer and makes HAE a more powerful and selective approach for large plant tissues, overcoming limitations commonly observed in ultrasound extraction.

Regarding the solvent and extraction time, both were decisive factors, as shown by the RSM plots (Figure 1). Total solids were strongly influenced by solvent polarity, with water providing higher solubilization of polar molecules. Flavonoids, particularly glycosides, dissolve readily due to their sugar moieties, abundant hydroxyl groups, and association with polysaccharides. Moreover, water’s high dielectric constant promotes extensive hydrogen bonding with polar compounds, enhancing extraction efficiency. Consequently, water proved more effective than ethanol for recovering these macromolecules (Plaskova and Mlcek, 2023; Tlili et al., 2026).

The results indicate that 50% ethanol provided the best overall extraction performance, achieving the highest TFC (23.5 µg QE/mg, S12) and maximum TS (23.8 mg/mL, S3). This behavior reflects the significant positive interaction between solvent composition and extraction time (Figure 1). Hydroalcoholic mixtures enhance metabolite recovery because ethanol disrupts hydrogen bonding while water interacts with the hydrophilic regions of the plant matrix (Plaskova and Mlcek, 2023). During the preliminary stage, this interaction was statistically significant (p = 0.005), with optimal performance at 15 min. Although highly polar solvents enable faster extraction, hydroalcoholic systems require longer infiltration time, which improves polyphenol solubilization and overall extraction selectivity (Sai-Ut et al., 2023).

Extraction time is a key factor during the process, as it controls the diffusion of bioactive compounds from the plant matrix into the extraction solvent, thereby maximizing yield (Sai-Ut et al., 2023). Although the interaction between residue type and extraction time was not statistically significant, the combination of these two factors demonstrated the most favorable numerical results for TS, the DPPH assay, and TFC when dried residue and a 15 min extraction time were applied. Under these conditions, TS ranged from 13.30 to 15.73 mg/mL, the DPPH assay reached 90% (extracts S10 and S11), and the maximum TFC was 23.5 µg QE/mg (extract S12) (Table 1). Wang et al. (Wang et al., 2018), while working on the optimization of extraction from Olea europaea leaves, reported higher flavonoid extraction efficiency at 50 min compared to 30 min for UAE, reaching a peak of 74.95 mg/g RE.

It is noteworthy that, although the solvent was critical for achieving greater flavonoid extraction efficiency, it had little relevance for the antioxidant activity. Therefore, this antioxidant activity may be associated with other bioactive compounds in addition to flavonoids. The work proposed by dos Santos et al. (dos Santos et al., 2025) reported the presence of several polyphenols, in addition to flavonoids, in the peel of Spondias tuberosa, such as coumarins, anthocyanidins, and tannins, as well as the presence of ascorbic acid with recognized antioxidant capacity (Ribeiro et al., 2019).

In summary, the preliminary stage demonstrated that while the extraction method itself was not statistically significant, HAE showed a strong numerical tendency toward superior overall efficiency, achieving the highest TS (23.8 mg/mL, extract S3) and a satisfactory flavonoid extraction efficiency (18.31 µg QE/mg, extract S3), with antioxidant activity reaching 90%. Therefore, based on these results, HAE was selected and fixed for the optimization stage due to the individual statistical significance of extraction time and solvent concentration, as well as their interaction (p < 0.05). Thus, the optimization was centered on adjusting these two parameters.

4.2 Optimization of the extraction process of Spondias tuberosa residues

The optimization of the extraction process maintained the same response variables as the screening stage: TS, TFC, and the DPPH assay.

Upon observing the RSM (Figure 2), the TS surface exhibited a “saddle” shape, in which the extremes achieved most favorable results. As illustrated by the Pareto chart (Figure 2), the linear effect of extraction time was negative, indicating an inverse relationship whereby increasing extraction time decreased total solids recovery. In contrast, the quadratic effects of extraction time and solvent concentration suggest the presence of an optimal extraction region, located around the center points of the experimental design. Additionally, less polar solvents require longer extraction times to enable diffusion and increased contact with the cellular matrix, unlike more polar solvents, which exhibit faster penetration (Sai-Ut et al., 2023).

According to Oreopoulou et al. (Oreopoulou et al., 2019), ethanol is more effective at solubilizing glycosides present in flavonoids, whereas water, in turn, is more capable of dissolving glycosides of phenolic acids. Therefore, binary solvents exhibit greater efficiency. Ribeiro et al. (Ribeiro et al., 2022) reported that the optimal solvent for compound recovery from Spondias tuberosa residue was 37% ethanol with a solid–liquid ratio of 1:38. In agreement, the same authors (Ribeiro et al., 2020) demonstrated that the best conditions for the recovery of secondary metabolites and antioxidants from the peel of Spondias purpurea L. employed 37% ethanol and a solid–liquid ratio of 1:33. Both assays are consistent with the present work, in which the highest TS was achieved using 30% ethanol, with a yield of 32.71 mg/mL (extract O2).

TFC showed statistically significant linear effects regarding extraction time and solvent concentration. Both factors had a positive impact on flavonoid recovery, as demonstrated by the Pareto chart (Figure 2). TFC increased with higher ethanol concentrations and extraction times, as illustrated by the RSM (Figure 2), reaching 29.76 µg QE/mg at 70% ethanol and 15 min. Similar solvent strength was reported by Oreopoulou et al. (Oreopoulou et al., 2019) for polyphenol recovery in complex raisin matrices. In contrast, Ribeiro et al. (Ribeiro et al., 2022) observed an optimal ethanol range of 20%–60% (peak at 55%) during thermally assisted extraction (TAE) of umbu peels. This divergence suggests that HAE overcame TAE limitations, as shear forces enhanced penetration into the fibrous matrix, improving solubilization of the less polar flavonoids without the saturation or thermal degradation reported for TAE. Moreover, Less polar solvents under mild conditions tend to reduce phenolic degradation during prolonged extraction (Oreopoulou et al., 2019). Similarly, Aparnadevi and Gokhale (Aparnadevi and Gokhale, 2026) reported a TFC of 133.5 mg QE/100 g DM after 60 min of conventional extraction of Vitis vinifera residue.

Regarding the DPPH assay, lower ethanol concentrations favor the recovery of antioxidant bioactive compounds. According to the RSM (Figure 2), ethanol concentrations between 50% and 30% exhibit antioxidant activity above 90%. This result is consistent with the assay proposed by Ribeiro et al. (Ribeiro et al., 2022), in which the optimized extraction condition employed 37% ethanol.

During the optimization stage, DPPH assay showed a significant quadratic effect (p = 0.026), indicating the existence of an optimal extraction time, as evidenced by the parabolic behavior of the RSM (Figure 2). A similar result was observed by Che Sulaiman et al. (Che Sulaiman et al., 2017), who optimized the reflux extraction of Clinacanthus nutans leaves by varying time, the water-to-ethanol, and temperature, obtaining a maximum activity of 76.25% at 80 min and a 90:10 solvent ratio. In comparison, the umbu residue exhibited higher activity (84.20%–95.56%) at drastically shorter times (2.9–17.1 min). In both studies, prolonged exposure to the extraction process led to a reduction in antioxidant activity.

Regarding the descriptive capacity of the optimization phase, the ordinary coefficients of determination for TS, TFC, and DPPH were consolidated at 61.96%, 51.32%, and 51.97%, respectively, with corresponding adjusted coefficients of 45.66%, 39.15%, and 46.64%. Crucially, while the individual main and quadratic effects for TS and TFC were found to be statistically significant, a statistically significant lack of fit was concurrently observed for both models (p = 0.010073 for TS and p = 0.020618 for TFC).

These two specific models must be interpreted as descriptive tools mapping directional extraction trends rather than as robust, independent predictive tools. In global multiresponse optimization, this significant lack of fit is a critical limitation that prevents the strict mathematical algorithms from operating as absolute predictors for TS and TFC. Consequently, these models cannot be evaluated in isolation; instead, they must be integrated into a comprehensive global assessment alongside the fully validated DPPH model (p_{lack of fit} = 0.203,731) and solid process engineering criteria, shifting the final selection toward a strategic operational compromise rather than an unmoderated algorithm-driven maximum.

This high variability is directly related to the inherent complexity of the umbu residue matrix. One of the most distinctive characteristics of fruits from the Spondias genus is the presence of a fibrous and sclerenchymatous endocarp. Consequently, the coexistence of rigid and soft tissues results in a complex and heterogeneous cell wall structure (Herrera et al., 2018).

In conclusion, the simultaneous optimization of the multiresponse extraction framework required a strategic technological trade-off, which was mathematically modeled using Derringer’s desirability function (Supplementary Figure S7). The strict mathematical algorithm identified an optimum at 21.72% ethanol and 10 min, corresponding to an overall desirability score of 0.99296. However, to bridge the gap between the mathematical modeling and practical chemical engineering reality, the final processing parameters were established by the authors at 50% ethanol (v/v) and 10 min, designated as the selected operational compromise condition.

Experimentally, the extracts produced at the center point (50% ethanol) demonstrated a superior TFC content TFC and enhanced DPPH scavenging activity compared to the runs conducted at 21.72% ethanol. Although the TS yield was mathematically lower at 50% ethanol, the high functional bioactivity (DPPH) heavily favored the hydroethanolic center point. Furthermore, from a downstream processing perspective, operating at 21.72% ethanol matrix generates an extract containing nearly 80% water. Evaporating this highly aqueous phase required a longer residence time in the rotary evaporator during laboratory testing. Even when conducted under mild vacuum and safe temperatures (<45 °C), this prolonged exposure to residual thermal stress poses a strict risk of degrading sensitive secondary metabolites and severely hinders industrial process scale-up.

This balanced operational compromise ensures a high-value-added extract with optimal antioxidant efficacy and robust manufacturing reproducibility, making it a viable, high-performing ingredient for the green cosmetics industry. The analysis of the solvent profile (Supplementary Figure S7) reveals that the response curves form a stable thermodynamic plateau between 21% and 50% ethanol, where the partial derivatives of the extraction yields with respect to solvent concentration approach zero, making the yield differences statistically non-significant.

Additionally, future studies will focus on the phytochemical characterization of the extracts through the application of advanced analytical techniques, such as liquid chromatography–mass spectrometry (LC–MS), with the aim of identifying and mapping the bioactive compounds present. Furthermore, multiple in vitro antioxidant assays will be performed to investigate the antioxidant mechanism of action throughout the oxidative cascade. Finally, further scale-up studies are required to confirm the feasibility of the extraction process and support its industrial application.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fntpr.2026.1887798/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Vanessa G Pasqualotto Severino, Universidade Federal de Goiás, Brazil

Reviewed by: Imen Zaghbib, Sevilla University, Spain

Kaio Vinicius Lira Da Silva Bastos, University of São Paulo, Brazil