Integrating Machine Learning and SHAP Analysis to Boost Callus Growth and Rutin Biosynthesis in Capparis spinosa L.

Integrating Machine Learning and SHAP Analysis to Boost Callus Growth and Rutin Biosynthesis in Capparis spinosa L. | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Integrating Machine Learning and SHAP Analysis to Boost Callus Growth and Rutin Biosynthesis in Capparis spinosa L. Marouane Mohaddab, Younes EL Goumi, Mohammed Elakrouch, Soufiane Hasni, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8619922/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Capparis spinosa L. is a Mediterranean medicinal species of high economic value, yet its large-scale propagation and metabolite production remain constrained by conventional approaches. A full factorial design was used to evaluate the effects of four plant growth regulators, 6-benzylaminopurine, kinetin, 2,4-dichlorophenoxyacetic acid, and 1-naphthaleneacetic acid, on fresh weight gain from leaf explants. Data from twenty hormonal treatments were modeled using four machine learning algorithms: Random Forest, Gradient Boosting, Extreme Gradient Boosting, and second-degree polynomial regression. Random Forest provided the highest predictive accuracy. SHapley Additive exPlanations analysis identified 2,4-dichlorophenoxyacetic acid as the dominant factor driving callogenesis, with 6-benzylaminopurine exerting a secondary synergistic effect, whereas kinetin and 1-naphthaleneacetic acid showed minimal or inhibitory influence. Experimental validation confirmed the five best Random Forest–predicted hormonal combinations, including the optimal mixture of 6-benzylaminopurine and 2,4-dichlorophenoxyacetic acid, which produced the highest increase in callus fresh weight gain. Rutin, the major bioactive flavonoid of C. spinosa , was identified by LC-QTOF-MS/MS tandem mass spectrometry and semi-quantified by LC-TQ-MS/MS under 6-benzylaminopurine and 2,4-dichlorophenoxyacetic acid combinations. A stacked Random Forest model integrating fresh weight gain predictions successfully estimated rutin accumulation, with maximal production at moderate hormone levels. This integrative machine learning and SHapley Additive exPlanations framework offers an interpretable and scalable strategy for optimizing callus culture and enhancing high-value metabolite production in C. spinosa . Moreover, callus culture represents a promising and sustainable alternative for large-scale production of valuable metabolites, reducing reliance on wild plant resources. Capparis spinosa L. Plant tissue culture Callogenesis Machine Learning Rutin Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Capparis spinosa L., an endemic species of arid Mediterranean regions, has attracted increasing scientific attention due to its medicinal, nutritional, and agroecological values. The species is extensively harvested and commercially exploited across several Mediterranean countries, including Morocco, Spain, Italy, Greece, and Turkey (Sozzi and Vicente 2006 ; Grimalt et al. 2018 , 2022 ; Hazrati et al. 2025 ). C. spinosa is characterized by a rich phytochemical profile, encompassing glucosinolates, phenolic acids, alkaloids, and flavonoids, which are associated with a wide range of biological activities such as antioxidants, antidiabetic, anti-inflammatory, and anticancer effects (Moghadamnia et al. 2019 ; Esmaeilpour et al. 2020 ; Rahimi et al. 2020 ). These properties highlight its considerable potential for applications in the pharmaceutical, nutraceutical, and cosmetic industries.(Karous et al. 2021 ; Kirkan et al. 2021 ; Annaz et al. 2022 ). Among these compounds, rutin stands out for its high economic and therapeutic value, making it a compound of particular interest (Fotiadou et al. 2025 ; El Amri et al. 2025 ). The biosynthesis of rutin involves branched metabolic pathways from phenylalanine and is mediated by phenylalanine ammonia-lyase (PAL) and other key enzymes (Kianersi et al. 2020 ). The exploitation of this plant supports the livelihoods of rural and marginalized communities while contributing to local economies (Sozzi et al. 2012 ). However, commercial harvesting relies mainly on wild populations, increasing pressure on natural resources and threatening long-term sustainability (Zhang and Ma 2018 ). Due to its tolerance to water deficit and adverse soils, C. spinosa also represents an ideal candidate for climate-resilient agriculture (Yousefi et al. 2025 ). Despite these advantages, traditional propagation methods, whether sexual or vegetative, face significant limitations, including low efficiency, slow growth, and the potential loss of desirable traits (Gristina et al. 2014 ). Pretreatments such as scarification, sulfuric acid exposure, or hormone application partially improve germination and rooting but remain insufficient for rapid large-scale multiplication (Labbafi et al. 2018 ; Nowruzian and Aalami 2023 ). In vitro culture techniques, particularly callus induction, offer an effective solution (Efferth 2019 ). They enable the regeneration of uniform, pathogen-free plants and may constitute effective reservoirs for the biosynthesis of secondary metabolites, thereby contributing to the conservation of natural populations and meeting the growing demand for phytotherapeutic resources (Yin et al. 2014 ; Mohaddab et al. 2022 ). Although research on C. spinosa remains limited (Caglar et al. 2005 ; Movafeghi et al. 2008 ; Al-Safadi and Elias 2011 ; Carra et al. 2012 ; Fahmideh et al. 2019 ; Gianguzzi et al. 2019 ; Mohaddab et al. 2024 ), callus tissue, being totipotent, serves as an ideal platform for targeted secondary metabolite production. Hagaggi et al. (2024) reported that callus extracts contain higher levels of phenolic and flavonoid compounds, a greater diversity of volatile metabolites, and superior antioxidant and antibacterial activities compared to leaf extracts. Thus, strict control of culture conditions enhances the biosynthesis of bioactive flavonoids, offering direct potential for industrial and pharmaceutical applications (Wijerathna-Yapa et al. 2025 ). Callus formation and proliferation depend on intrinsic factors, including explant type, age, and medium composition, as well as extrinsic factors such as light, temperature, and humidity (Pasternak and Steinmacher 2024 ). Plant growth regulators (PGRs), particularly auxins and cytokinins, are essential for morphogenesis, biomass accumulation, and secondary metabolite production (Ikeuchi et al. 2013 ; Bravo-Vázquez et al. 2023 ). Commonly used phytohormones include 2,4-dichlorophenoxyacetic acid (2,4-D) and 1-naphthaleneacetic acid (NAA), applied alone or in combination with cytokinins such as 6-benzylaminopurine (BAP) or kinetin (KIN) (Koufan et al. 2022 ; Higazy et al. 2023 ). Studies indicate that 2,4-D at 1.0 mg/L combined with BAP at 1.5 mg/L effectively induces callus from C. spinosa leaf explants (Wang et al. 2007 ; Liu et al. 2011 ; Yin et al. 2014 ). Strategic manipulation of PGRs during callogenesis can therefore enhance both callus regeneration and flavonoid production, particularly rutin (Sobhy et al. 2025 ). In vitro plant development is a complex process governed by multiple interconnected variables (Ma et al. 2018 ; Svolacchia and Sabatini 2023 ). Studying these factors in isolation is insufficient, as their interactions strongly influence outcomes. Classical statistical approaches are limited when handling nonlinear and multidimensional datasets and often exhibit high error rates (Aasim et al. 2023 ; Katırcı et al. 2024 ). Mechanistic models, although grounded in biological knowledge, require precise measurement of numerous parameters that are difficult to standardize experimentally (Noordijk et al. 2024 ), reducing their predictive power and practical usefulness for rapid protocol optimization (Baker et al. 2018 ). Artificial intelligence (AI), especially supervised Learning or Machine Learning (ML), offers a means to manage this complexity (Peng and Rajjou 2024 ). ML applications to callogenesis and organogenesis represent an emerging and promising field (Hesami and Jones 2021 ). Algorithms such as Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost) have been successfully applied to predict callus induction, shoot regeneration, biomass yield, and secondary metabolite production in various species (Hesami and Jones 2021 ; Alcalde et al. 2022 ; Eren et al. 2023 ; Demirel et al. 2023 ; Sarabandi et al. 2024 ; Bozkurt et al. 2024 ; Tan et al. 2024 ; Zarbakhsh et al. 2024 ; Koçak et al. 2025 ; Sarmah et al. 2025 ). Despite their predictive performance, these models often remain “black boxes,” where relationships between biological variables and observed responses are opaque (Petch et al. 2022 ). SHapley Additive exPlanations (SHAP) values provide a solution to this opacity. Derived from game theory, they quantify the contribution of each variable to the final prediction by considering all possible factor combinations (Lundberg and Lee 2017 ; Lundberg et al. 2020 ). By transforming a black-box model into a “white-box” model, SHAP values allow identification of key variables and their interactions, offering biologically meaningful interpretation (Ekanayake et al. 2022 ). Although interest in C. spinosa is increasing, investigations into callus culture and flavonoid biosynthesis, particularly rutin, remain limited and fragmented. The majority of studies have focused on isolated factors rather than the complex, non-linear interactions among growth regulators, explant type, and environmental conditions (Dar et al. 2021 ). Furthermore, while machine learning approaches have demonstrated the capacity to drive diverse tissue culture systems, their application to C. spinosa , especially using interpretable methods such as SHAP, has yet to be explored, leaving critical variables governing flavonoid production largely unexplored. This study examined the effects of growth regulators, individually and in combination, on callus induction and proliferation, as well as on rutin biosynthesis. The collected data were evaluated using machine learning models, complemented by SHAP analyses, to quantify the relative influence of each regulator and identify the key factors modulating both callus development and rutin biosynthesis. Materials and methods Collection of Plant Material Fresh leaves of C. spinosa L. were collected in April 2022 from the Taounate region (Morocco; latitude: 34°26'31.187" N, longitude: 4°42'11.97" W, altitude: 435 m). The leaves were harvested on the same day they were processed for further experiments. Surface Sterilization Harvested leaves were surface sterilized prior to culture. Sterilization was performed following the protocol described by Slimani et al. ( 2021 ), ensuring that explants were free from microbial contamination before placement on culture medium. Culture Medium Preparation Callus induction was carried out on Murashige and Skoog (MS) medium ( Murashige and Skoog 1962 ), supplemented with 0.8% agar, 3% sucrose, and MES buffer (M0255, Duchefa). Leaf explants, approximately 1 cm in diameter, were placed abaxial side down on 20 mL of medium in each Petri dish, with five explants per dish. A total of 1,000 explants were used in the experiment. The pH of the medium was adjusted to 5.7 ± 0.1 before autoclaving at 121°C for 20 minutes under 1 bar of pressure. Plant Growth Regulator Combination Design The basal medium was supplemented with different concentrations of exogenous plant growth regulators (PGRs) to evaluate their effects on callus induction. The PGRs tested included 2,4-dichlorophenoxyacetic acid (2,4-D: 0.0–2.0 mg/L), 6-benzylaminopurine (BAP: 0.0–2.0 mg/L), 1-naphthaleneacetic acid (NAA: 0.0–2.0 mg/L), and kinetin (KIN: 0.0–2.0 mg/L). A full factorial design was applied with four PGR combinations: BAP/NAA, BAP/2,4-D, KIN/NAA, and KIN/2,4-D. For each combination, four concentrations (0.0–1.5 mg/L) were applied per regulator, resulting in 16 unique concentration pairs per combination. Each PGR was also tested individually at 2.0 mg/L, for a total of 20 distinct treatments, each replicated three times. In vitro culture Conditions All cultures were maintained in a phytotron under a 16-hour light/8-hour dark photoperiod at 24 ± 2°C for 30 days. Illumination was provided by Sylvania fluorescent tubes delivering a light intensity of approximately 1600 W. Measurement of Callus Growth After 30 days, callus Fresh Weight Gain (FWG) was measured by weighing callus immediately after removal from the medium. These data were then compiled into a dataset for machine Learning analysis. FWG was calculated as: \(\:\text{F}\text{W}\text{G}=\:\text{F}\text{i}\text{n}\text{a}\text{l}\:\text{f}\text{r}\text{e}\text{s}\text{h}\:\text{w}\text{e}\text{i}\text{g}\text{h}\text{t}\:-\:\text{I}\text{n}\text{i}\text{t}\text{i}\text{a}\text{l}\:\text{w}\text{e}\text{i}\text{g}\text{h}\text{t}\:\text{o}\text{f}\:\text{e}\text{x}\text{p}\text{l}\text{a}\text{n}\text{t}\) Identification and semi-quantification of rutin in callus Cultures using LC–MS/MS Approximately 150 mg of bone ash, dried at 30°C to constant weight, was extracted according to the protocol of Belkessam et al. ( 2025 ), with minor modifications, using 2 mL of methanol–water (80:20, v/v) and 5 µL of flavone as an internal standard. The samples were sonicated for 20 minutes, centrifuged for 10 minutes, and then filtered through 0.22 µm filters before analysis by LC–MS/MS. Rutin was qualitatively identified by LC-QTOF and semi-quantified by LC-TQ via the MRM transition 609 → 300, in accordance with the original method. Acquisition and processing of the data were carried out with MassHunter® Workstation software (version 10.0, Agilent Technologies, Santa Clara, CA, USA). $$\:\text{M}\text{o}\text{l}\text{e}\text{c}\text{u}\text{l}\text{e}\:\text{c}\text{o}\text{n}\text{t}\text{e}\text{n}\text{t}\:\left(\text{A}.\text{U}./mg\:\text{D}\text{W}\right)=\frac{\text{M}\text{R}\text{M}\:\text{p}\text{e}\text{a}\text{k}\:\text{a}\text{r}\text{e}\text{a}\:\text{o}\text{f}\:\text{r}\text{u}\text{t}\text{i}\text{n}}{\text{M}\text{R}\text{M}\:\text{f}\text{l}\text{a}\text{v}\text{o}\text{n}\text{e}\:\text{p}\text{e}\text{a}\text{k}\:\text{a}\text{r}\text{e}\text{a}\:\text{x}\:\text{S}\text{a}\text{m}\text{p}\text{l}\text{e}\:\text{d}\text{r}\text{y}\:\text{w}\text{e}\text{i}\text{g}\text{h}\text{t}\:}$$ A.U. = Arbitrary Units; DW = Dry Weight ; MRM = Multiple Reaction Monitoring. Modeling procedures The study methodology is presented in figure. 1. Callogenesis was induced using four PGR combinations: BAP/NAA, BAP/2,4-D, KIN/NAA, and KIN/2,4-D at varying concentrations to evaluate their effects on FWG and rutin accumulation. To ensure data quality, the data were normalized, outliers were removed, and the relationships between parameters were examined using a Pearson correlation test. The final dataset was partitioned into a training set (75%) and an independent test set (25%), while maintaining a balanced distribution of experimental conditions (Figure. 1 (A)). Figure 1 A schematic representation of the stepwise computational approach used in this study. (A) Modeling and prediction of callus Fresh Weight Gain (FWG) (output) based on 6-Benzylaminopurine (BAP), Kinetin (KIN), Naphthalene Acetic Acid (NAA), and 2,4-Dichlorophenoxyacetic Acid (2,4-D) as inputs using four models. (B) SHAP value analysis of the best-selected model to identify and quantify the key model features. (C) Stacking of the best model and using it as input for the production and accumulation of rutin (Model 2). Four machine learning models, RF, GB, XGBoost, and second-degree polynomial regression (PL2), were evaluated for FWG prediction. Model performance was assessed using established statistical indicators: the coefficient of determination (R²) and cross-validation (R²CV), moreover root mean square error (RMSE). Low RMSE values indicate that the model’s predictions are reliable, with minimal error. R² quantifies the proportion of the total variance in the observed data captured by the model, while R²CV assesses its ability to generalize to new data. Both range from 0 to 1, with values closer to 1 reflecting strong agreement between predicted and observed values and good model stability. Overall, a model with high R² and R²CV combined with low RMSE can accurately and reliably reproduce the behavior of the studied system. SHAP effectively serves as a sensitivity analysis (Fig. 1 (B)), measuring the impact of each growth regulator on FWG while accounting for the combined effects of all other variables. This approach allows the identification of regulators with the greatest influence on callus growth and provides insight into the complex interactions among them. Predictions from model 1 were then incorporated into a stacked model designed to predict rutin accumulation (model 2) (Fig. 1 (C)). This model used both the initial concentrations of the growth regulators and the FWG predictions as input variables, directly linking callus growth to secondary metabolite production. To validate the efficacy and reliability of this optimized model, the predicted conditions for maximizing FWG and rutin accumulation were experimentally tested in triplicate. The results confirmed the accuracy of the predictions and demonstrated the practical utility of the model for optimizing callogenesis and secondary metabolite production. Results Influence of plant growth regulators on C. Spinosa callus Induction This study investigated the influence of phytohormones matrix at different concentrations (0, 0.5, 1 and 1.5 mg/L) and hormonal combinations (BAP/NAA), (BAP/2,4-D), (KIN/NAA), and (KIN/2,4-D) on the FWG of callus derived from C. spinosa leaf explants after 30 days of culture (Table A). Notably, no callus formation occurred in the culture medium lacking PGRs, highlighting the essential role of exogenous hormones in inducing callogenesis in C. spinosa leaf explants. Among the tested hormone combinations, 2,4-D and BAP showed an effect on callus induction and growth. The highest FWG (3639.5 mg ± 208.0) was recorded with a combination of 1 mg/L BAP with 1 mg/L 2,4-D. Similarly high values were observed with 1 mg/L BAP + 0.5 mg/L 2,4-D (3459.33 mg ± 84.68). Additionally, 2,4-D applied alone also promoted substantial callus proliferation, with FWG values of 3438.5 mg ± 493.5 and 3394 mg ± 31 at 0.5 and 1 mg/L, respectively. In contrast, the (BAP/ NAA) and (KIN/ NAA) combinations produced significantly lower FWG, indicating less efficiency in promoting callus formation and proliferation. The maximum FWG in these groups was observed with 0.5 mg/L BAP combined with 1.5 NAA (1874.45 mg ± 25.66) and with 0.5 mg/L KIN combined with 1 mg/L NAA (1350.5 mg ± 215.5). To further investigate these complex hormonal interactions and their effects on callus induction, artificial intelligence was employed to analyze the dataset. Model performance in predicting callus induction The callogenesis of C. spinosa was quantitatively modeled using a set of explanatory variables representing four exogenously applied PGRs: BAP, KN, NAA, and 2,4-D. The target variable for modeling was the FWG of callus tissue. To capture the complex relationships between hormonal treatments (inputs) and FWG (outputs), four ML models were applied and compared: RF, GB, XGBoost, and a traditional second-degree Polynomial Regression. Table 1 Machine learning model performance for predicting callus fresh weight gain. ML Models R 2 R 2 CV RMSE RF 0.86 0.89 263.33 GB 0.83 0.86 285.59 XGBoost 0.85 0.86 274.46 PL2 0.56 0.63 471.98 R²: coefficient of determination, R²CV: cross-validated R², RMSE: root mean square error, RF: Random Forest, GB: Gradient Boosting, XGBoost: Extreme Gradient Boosting, PL2: second-degree polynomial regression. The ensemble Learning models, RF, GB, and XGBoost, demonstrated superior and comparable performance (Table 1 ). RF had the highest predictive accuracy with an R² of 0.86, R²CV of 0.89, and a low RMSE of 263.33, indicating precise predictions and strong generalization. XGBoost and GB also performed well, with R² values of 0.85 and 0.83, R²CV of 0.86 for both, and RMSE values of 274.46 and 285.59, respectively. These results highlight the ability of ensemble methods to capture complex, non-linear relationships in FWG data effectively. In contrast, the PL2 model exhibited substantially lower performance, with an R² of 0.56, R²CV of 0.63, and a markedly higher RMSE of 471.98. This indicates that simple polynomial relationships are insufficient to model the underlying biological complexity, resulting in larger prediction errors. Overall, the findings suggest that ensemble learning approaches, particularly RF and XGBoost, provide reliable and robust predictions for FWG. Figure 2 Scatter plots presenting the values of real vs predictions of callus fresh weight gain for four models: Random Forest, Gradient Boosting, XGBoost, and second-degree Polynomial regression. The combined analysis of numerical and graphical results highlights the superiority of ensemble tree-based models (RF, GB, and XGBoost) for predicting callus FWG. As shown in Fig. 2 , these models exhibit regression slopes close to unity (RF: 0.98 x + 42.04; GB: 0.94 x + 83.13; XGBoost: 0.96 x + 71.62) and a tight clustering of points around the ideal y = x line, indicating precise calibration and limited residual dispersion. These findings are consistent with the high coefficients of determination and low cross-validation errors (R²CV) reported in Table 1 , confirming the robustness and generalizability of these models across the full FWG range. In contrast, PL2 shows a markedly lower slope (0.81 x + 244.11) and substantial dispersion, reflecting systematic underestimation of high FWG values and poor ability to capture the non-linearities inherent in biological data. Overall, these results indicate that ensemble models, particularly Random Forest, represent the most accurate and reliable approach for predicting callus fresh weight gain, whereas polynomial regression remains unsuitable for modeling such complex relationships. The graphs show the overall importance of the variables (Figure. 2A) reveal that 2,4-D has the highest absolute mean SHAP value (~ 700), significantly surpassing those of BAP, KIN, and NAA, making it the most influential factor in predicting callus biomass accumulation. Analyzing the meaning of SHAP values (Figure. 3B), 2,4-D has a strongly positive effect on FWG (+ 12.19). Comparatively, BAP also contributes positively, but more moderately (+ 6.39). In contrast, KIN reduces predicted growth (-6.95), and NAA has a slightly negative impact (-2.01), highlighting its marginal role. Thus, these results confirm that 2,4-D is the key regulator of callus growth, BAP plays a secondary favorable role, while KIN may act as an inhibitor, and NAA remains of little influence. Figure 3 Global importance (A) and directional effects (B) of hormones on predicted callus FWG using SHAP values in Random Forest models. 6-benzylaminopurine (BAP), kinetin (KIN), naphthaleneacetic acid (NAA), and 2,4-dichlorophenoxyacetic acid (2,4-D). Experimental validation of callus fresh weight gain Laboratory experiments were conducted to validate the five top phytohormone combinations predicted by the RF model and assess the reliability of the results. The validation experiment results (Figure. 4) demonstrated the reliability and accuracy of RF in predicting optimal combinations of plant growth regulators, enabling the achievement of maximal callus FWG. Results from the validation experiment provide robust evidence for the effectiveness of decision tree models, particularly Random Forest, as a powerful tool for optimizing callus induction. This approach also paves the way for suspension culture and large-scale production, while further enabling the optimization of micropropagation protocols for plants, such as vegetative propagation of C. Spinosa . Callus culture as a platform for rutin accumulation Hormonal regulation is a key factor in controlling the biosynthesis of secondary metabolites in C. spinosa callus cultures. In this study, rutin, the main metabolite of this species, was first identified using UPLC-QTOF-MS/MS, while relative concentration was then semi-quantified using UPLC-TQ-MS/MS in MRM mode (609 → 300). The BAP/2,4-D hormone combination was chosen for these experiments due to its favorable influence on FWG (Figure. 3), creating optimal conditions to assess rutin accumulation in callus tissues. The influence of hormonal treatments on rutin content was further analyzed using an RF model. Figure. 5 presents a comparison between experimentally determined and predicted rutin concentrations. The data indicate that rutin accumulation is strongly dependent on the relative concentrations of BAP and 2,4-D. Maximum rutin levels were observed under moderate BAP concentrations (0.5–1.0 mg/L) combined with low 2,4-D concentrations (0.0–1.0 mg/L), suggesting that these conditions favor activation of the flavonoid biosynthetic pathway. In contrast, higher concentrations of either BAP (≥ 1.5 mg/L) or 2,4-D (≥ 1.5 mg/L) markedly reduced rutin content, implying an inhibitory effect of excessive hormone levels on secondary metabolism. Overall, the RF model effectively captured these trends, with predicted values closely aligning with experimental observations. Figure. 6 further illustrates the predictive performance of the RF model. Observed versus predicted rutin concentrations cluster around the ideal 1:1 line, underscoring the model’s ability to faithfully reproduce the relationship between measured and estimated values. Regression analysis yielded a slope of 0.79 and an intercept of 0.25, with a high coefficient of determination (R² = 0.821) and a highly significant p -value, confirming the statistical robustness and reliability of the model. These results demonstrate that the model accurately captures overall trends, even within the complexity of a biological system. Ensemble tree-based algorithms are notable for their flexibility, capacity to model high-order nonlinear interactions, ability to handle heteroscedasticity, and robustness against overfitting. These qualities render the model a powerful tool for predicting and understanding the biosynthesis of secondary metabolites, such as rutin, in plant tissue cultures. Experimental validation of rutin accumulation in callus cultures The predictions of rutin accumulation obtained by the RF model, based on the BAP/2, 4-D combinations presented in section 3.3, show overall agreement with the experimental data, particularly for the condition BAP = 2.0 + 2.4-D = 2.0, highlighting the robustness of the model in anticipating metabolic responses to different hormonal treatments. The slight overestimation observed for certain combinations (BAP = 1.5 + 2.4-D = 2.0 and BAP = 2.0 + 2.4-D = 1.0) does not alter the overall consistency of the predictions. These results demonstrate that the model is a reliable tool for estimating rutin accumulation in C. spinosa callus cultures and provide effective support for optimizing culture conditions to maximize secondary metabolite production. Discussion Callus induction is a central step in plant tissue culture, representing a cellular reprogramming process in which differentiated cells revert to an undifferentiated and proliferative state (Lee et al. 2024 ). This process is primarily regulated by the exogenous application of PGRs, which provide the plasticity required for subsequent morphogenetic events, such as organogenesis and somatic embryogenesis (Pasternak and Steinmacher 2024 ). Callus also serves as a reservoir for secondary metabolite production, including rutin. Therefore, investigating the effects of hormones, cytokinins, and auxins on callus induction and rutin accumulation is crucial for effective utilization of this biological resource. In this study, we evaluated the effects of varying concentrations of BAP, KIN, NAA, and 2,4-D on FWG and rutin accumulation in C. spinosa using RF, GB, XGBoost, and PL2. ML algorithms, including RF and XGBoost, have been studied in plant tissue culture, yet comparative analyses remain limited. These findings highlight that classic statistical methods, based on linear relationships or limited curvature, struggle to capture the complex interactions, threshold effects, and saturation phenomena characteristic of plant hormonal responses (Sadat-Hosseini et al. 2022 ). Modeling callus formation in vitro , where multiple hormonal factors interact in a nonlinear and potentially synergistic manner, requires more flexible approaches (Fallah Ziarani et al. 2022 ). Machine learning models, particularly tree-based ensemble algorithms such as RF and XGBoost, have proved particularly well suited to this challenge due to their ability to model high-order interactions, handle heteroscedasticity, and remain robust to overfitting (Munasinghe et al. 2020 ; Ali and Aasim 2024 ; Kaushik et al. 2025 ). The demonstrated accuracy of these models highlights the value of integrating machine learning into plant tissue culture research, not only to improve predictive ability but also to improve our understanding of the complex mechanisms driving hormonal responses. To evaluate the relative contributions of individual growth regulators to callus development, we focused the SHAP analysis on the RF model, selected for its superior performance in terms of predictive accuracy. This approach allows for a quantitative assessment of the effect of each factor (BAP, KIN, NAA, and 2,4-D) on FWG predictions for C. spinosa callus tissue, while taking into account potential interactions between variables. Recent studies comparing machine learning models for modeling and predicting saffron ( Crocus sativus ) callogenesis demonstrated that XGBoost and GB achieved superior accuracy (R² > 0.95) compared to artificial neural networks and RF regression in modeling in vitro culture systems (Sarabandi et al. 2024 ). In our study, we compare the predictive performance of RF, GB, XGBoost, and PL2 for FWG in C. spinosa . Tree-based Learning models consistently outperformed polynomial regression, with RF achieving the highest predictive accuracy (R² > 0.86) for callogenesis. Although this represents the first direct comparison of these models in C. spinosa in vitro culture, other studies have similarly reported the superior accuracy of RF (Özcan et al. 2023 ; Bozkurt et al. 2024 ; Sarmah et al. 2025 ). Overall, a model’s predictive capability is largely determined by the size and complexity of the training dataset (Hesami and Jones 2020 ). In this context, the RF model has consistently demonstrated a remarkable ability to accurately predict complex callus-related biological outcomes, achieving a high coefficient of determination (R² = 0.97) for both callus growth parameters and secondary metabolite production in Nilgirianthus ciliatus (Jeevan Ram et al. 2025 ). These results confirm RF as a robust and reliable tool for predicting callus growth and metabolite accumulation, providing confidence in its application for optimizing in vitro culture conditions. Sensitivity analyses were applied to interpret the RF model and quantify the contribution of each variable to FWG predictions. SHAP values as an interpretative tool for identifying and quantifying variable importance have been demonstrated in multiple studies (Ekanayake et al. 2022 ; Antonini et al. 2024 ). SHAP provides a mathematically rigorous, objective, and consistent method to decompose model predictions into contributions from each variable (Lundberg and Lee 2017 ). This approach enables precise local explanation of individual predictions while offering a global perspective on variable importance, making it robust and broadly applicable to machine learning interpretation (Lundberg et al. 2020 ). Using SHAP, key variables such as 2,4-D, NAA, sucrose concentration, and culture duration have been successfully identified (Sarabandi et al. 2024 ). In our study, SHAP analysis confirmed that 2,4-D was the most influential factor, KIN exhibited inhibitory effects, and BAP acted synergistically with 2,4-D to enhance callus formation. Identification of these factors is particularly relevant when secondary metabolite production is the primary objective of callus culture. To our knowledge, this study provides the first systematic assessment of the impact of various PGRs on both callus induction and rutin production in C. spinosa . To directly link callus growth with metabolite production, we employed a stacked RF model, using callogenesis outputs as inputs to predict rutin accumulation. This approach effectively modeled the nonlinear relationship between cell proliferation and rutin biosynthesis, yielding high predictive performance (R² > 0.82) and minimal deviation between observed and predicted values, confirming method robustness. For instance, Yin et al. ( 2014 ) reported that MS medium supplemented with 1.5 mg/L 2,4-D and 3.0 mg/L BAP induces callogenesis and that callus may serve as a source of volatile organic compounds. Similarly, Duran and Issah ( 2022 ) demonstrated that BAP at 1 mg/L combined with 2 mg/L NAA produced a callogenesis rate of approximately 120 mg after 10 days. While these studies highlighted the impact of the strigolactone GR24 on callus production and phenolic content in C. spinosa , our results showed that BAP + NAA achieved over 1000 mg FWG over 30 days, although it was less effective than BAP + 2,4-D. This confirms that NAA is less favorable for callogenesis. This combination also resulted in substantial rutin accumulation, with a modest increase in the presence of 0.1 µM GR24, indicating that C. spinosa callus can serve as a rutin reservoir (Duran and Issah 2022 ). Direct comparison between studies is limited due to differences in explant source, methodology, and objectives. In the present work, maximum callus formation was observed with 2,4-D, consistent with reports in C. spinosa (Sobhy et al. 2025 ), and other species where auxin application significantly promotes callogenesis (Slimani et al. 2021 ; Teoh et al. 2023 ; Ranade et al. 2023 ; Mahood et al. 2024 ; Abdelazeez et al. 2025 ). Regarding rutin accumulation, low concentrations of BAP and 2,4-D (≤ 1 mg/L) promoted significant accumulation, peaking at 0.5 mg/L for each hormone, paralleling biomass increases. Similar trends were reported by Kianersi et al. ( 2020 ), with further enhancements achieved using elicitors such as methyl jasmonate. (Goda et al. 2017 ) also reported that 2,4-D-induced callus exhibited significantly higher total flavonoid content. These findings highlight the importance of optimizing hormone concentrations and combining them with elicitors for large-scale production of high-quality bioactive compounds (Sagharyan et al. 2020 ; Ozyigit et al. 2023 ; Pérez-Mejía et al. 2024 ). The reliability of RF predictions was validated experimentally. Predicted maximal FWG results closely matched observed values, demonstrating the model’s accuracy. For rutin accumulation, satisfactory performance was observed, particularly for BAP + 2,4-D at 2 mg/L, with minor overestimations in some combinations. These results confirm RF as a robust and effective machine learning algorithm for modeling and optimizing callus culture in this study. Future studies using larger datasets and additional variables could further improve model performance. Investigations on C. spinosa could explore 2,4-D with other PGRs or extend to cell suspension cultures and bioreactor-based production systems. Conclusion This research provides the first comprehensive evaluation of the combined effects of BAP, KIN, NAA, and 2,4-D on callogenesis and rutin biosynthesis in C. spinosa using an integrated machine learning and explainable-AI framework. Ensemble tree-based models, particularly Random Forest, accurately predicted callus biomass and metabolite accumulation, outperforming classical statistical approaches. SHAP analysis revealed 2,4-D as the primary determinant of callus proliferation, while BAP acted synergistically to enhance growth, and KIN and NAA contributed weak or inhibitory effects. Experimental validation confirmed the reliability of the predicted optimal hormonal combinations, demonstrating strong agreement between predicted and observed FWG and rutin levels. Moderate concentrations of BAP and 2,4-D favored rutin accumulation, underscoring the importance of fine hormonal tuning for metabolic optimization. Overall, this study highlights the relevance of ML-driven modeling for deciphering complex hormonal interactions and accelerating protocol optimization in plant tissue culture. The established models constitute a robust foundation for scaling up callus-based production systems, including suspension cultures and bioreactors, aimed at sustainable, high-yield biosynthesis of rutin and other valuable metabolites in C. spinosa . Future investigations integrating larger datasets, molecular markers, and elicitation strategies will further strengthen predictive performance and advance the biotechnological exploitation of this species. Declarations CRediT authorship contribution statement Marouane Mohaddab: Writing – review & editing, Writing – original draft, Conceptualization, Methodology, Data curation, Formal analysis. Younes EL Goumi: Supervision, Writing – review and editing, Methodology. Mohammed Elakrouch: Formal analysis, Software, Methodology Soufiane Hasni: Formal analysis, Software, Data curation. Clément Burgeon: Writing – review & editing, Software, Data curation. Manon Genva: Writing – review & editing, Software. Malika Fakiri: Writing – review & editing, Validation, Project administration Marie-Laure Fauconnier: Writing – review & editing, Validation , Project administration Ethics declaration Not applicable. Declaration of interests The authors declare that there is no conflict of interest. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. References Aasim M, Ali SA, Altaf MT et al (2023) Artificial neural network and decision tree facilitated prediction and validation of cytokinin-auxin induced in vitro organogenesis of sorghum ( Sorghum bicolor L). Plant Cell Tiss Organ Cult 153:611–624. https://doi.org/10.1007/s11240-023-02498-3 Abdelazeez WMA, Aboueldis GR, Suliman AA, Mohammed DM (2025) Production of secondary metabolites in callus cultures of Scutellaria baicalensis L. and assessment of their anti-inflammatory and antioxidant efficacy in ulcerative colitis rats. Plant Cell Tiss Organ Cult 160:80. https://doi.org/10.1007/s11240-025-02996-6 Alcalde MA, Müller M, Munné-Bosch S et al (2022) Using machine learning to link the influence of transferred Agrobacterium rhizogenes genes to the hormone profile and morphological traits in Centella asiatica hairy roots. Front Plant Sci 13:1001023 Ali SA, Aasim M (2024) Response surface methodology and artificial intelligence modeling for in vitro regeneration of Brazilian micro sword ( Lilaeopsis brasiliensis ). Plant Cell Tiss Organ Cult 157:10. https://doi.org/10.1007/s11240-024-02734-4 Al-Safadi B, Elias R (2011) Improvement of caper ( Capparis spinosa L.) propagation using in vitro culture and gamma irradiation. Scientia Horti 127:290–297 Annaz H, Sane Y, Bitchagno GTM et al (2022) Caper ( Capparis spinosa L.): an updated review on its phytochemistry, nutritional value, traditional uses, and therapeutic potential. Front Pharmacol. 2022; 13: 1–22 Antonini AS, Tanzola J, Asiain L et al (2024) Machine learning model interpretability using SHAP values: Application to Igneous Rock Classification task. Appl Comput Geosci 23:100178. https://doi.org/10.1016/j.acags.2024.100178 Baker RE, Peña J-M, Jayamohan J, Jérusalem A (2018) Mechanistic models versus machine learning, a fight worth fighting for the biological community? Biol Lett 14:20170660. https://doi.org/10.1098/rsbl.2017.0660 Belkessam M, Genva M, Kouki A et al (2025) Phytochemical Profile and Cosmeceutical Potential of Leaf Extracts of Two Species of the Anacardiaceae Family from the Mediterranean Scrubland: Pistacia lentiscus L. and Pistacia atlantica Desf. Processes 13:3712. https://doi.org/10.3390/pr13113712 Bozkurt T, İnan S, Dündar İ et al (2024) Optimizing the In Vitro Propagation of Tea Plants: A Comparative Analysis of Machine Learning Models. Horticulturae 10:721. https://doi.org/10.3390/horticulturae10070721 Bravo-Vázquez LA, Angulo–Bejarano PI, Bandyopadhyay A et al (2023) Regulatory roles of noncoding RNAs in callus induction and plant cell dedifferentiation. Plant Cell Rep 42:689–705. https://doi.org/10.1007/s00299-023-02992-0 Caglar G, Caglar S, Ergin O, Yarim M (2005) The influence of growth regulators on shoot proliferation and rooting of in vitro propagated caper. J Environ Biol 26:479–485 Carra A, Del Signore MB, Sottile F et al (2012) Potential use of new diphenylurea derivatives in micropropagation of Capparis spinosa L. Plant Growth Regul 66:229–237. https://doi.org/10.1007/s10725-011-9645-3 Dar SA, Nawchoo IA, Tyub S, Kamili AN (2021) Effect of plant growth regulators on in vitro induction and maintenance of callus from leaf and root explants of Atropa acuminata Royle ex Lindl. Biotechnol Rep 32:e00688 Demirel F, Uğur R, Popescu GC et al (2023) Usage of Machine Learning Algorithms for Establishing an Effective Protocol for the In Vitro Micropropagation Ability of Black Chokeberry ( Aronia melanocarpa (Michx.) Elliott). Horticulturae 9:1112. https://doi.org/10.3390/horticulturae9101112 Duran RE, Issah H (2022) The impact of strigolactone GR24 on Capparis spinosa L. callus production and phenolic compound content. Plant Cell Tiss Organ Cult 149:197–204 Efferth T (2019) Biotechnology applications of plant callus cultures. Engineering 5:50–59 Ekanayake IU, Meddage DPP, Rathnayake U (2022) A novel approach to explain the black-box nature of machine learning in compressive strength predictions of concrete using Shapley additive explanations (SHAP). Case Stud Constr Mate 16:e01059. https://doi.org/10.1016/j.cscm.2022.e01059 El Amri N, Chabir R, Kachkoul R, Bour A (2025) Phenolic Composition and Phytochemical Analysis of Moroccan Raw and Processed Capparis spinosa Buds and Berries. Chem Biodivers. https://doi.org/10.1002/cbdv.202500443 Eren B, Türkoğlu A, Haliloğlu K et al (2023) Investigation of the influence of polyamines on mature embryo culture and DNA methylation of wheat ( Triticum aestivum L.) using the machine learning algorithm method. Plants 12:3261 Esmaeilpour HR, Boostani R, Shoeibi A et al (2020) The potential effects of Caper ( Capparis spinosa L.) in the treatment of diabetic neuropathy. Trad Med Res 5:442–448 Fahmideh L, Sheikhi M, Benakashani F, Solouki M (2019) Callus induction and organogenesis from various explants of plant Capparis spinosa L. under In vitro conditions. J Plant Prod Res 26:75–88. https://doi.org/10.22069/jopp.2019.14326.2284 Fallah Ziarani M, Tohidfar M, Navvabi M (2022) Modeling and optimizing in vitro percentage and speed callus induction of carrot via Multilayer Perceptron-Single point discrete GA and radial basis function. BMC Biotechnol 22. https://doi.org/10.1186/s12896-022-00764-4 Fotiadou E, Cheilari A, Graikou K, Chinou I (2025) Comparative Study by HPTLC of Selected Capparis spinosa Samples (Buds and Leaves) from the Cycladic Islands in Greece. Foods 14:1827 Gianguzzi V, Inglese P, Barone E, Sottile F (2019) In Vitro Regeneration of Capparis spinosa L. by Using a Temporary Immersion System. Plants 8:177. https://doi.org/10.3390/plants8060177 Goda SM, Ahmed SA, El Sherif F et al (2017) Genetically stable plants with boosted flavonoids content after in vitro regeneration of the endangered Capparis spinosa L. Glob. Drugs Ther 2:1–7 Grimalt M, Hernández F, Legua P et al (2018) Physicochemical composition and antioxidant activity of three Spanish caper ( Capparis spinosa L.) fruit cultivars in three stages of development. Scientia hortic 240:509–515 Grimalt M, Hernández F, Legua P et al (2022) Antioxidant activity and the physicochemical composition of young caper shoots ( Capparis spinosa L.) of different Spanish cultivars. Scientia Hortic 293:110646 Gristina AS, Fici S, Siragusa M et al (2014) Hybridization in Capparis spinosa L.: Molecular and morphological evidence from a Mediterranean island complex. Flora 209:733–741 Hagaggi NShA, Abdul-Raouf UM, Radwan TAA (2024) Variation of antibacterial and antioxidant secondary metabolites and volatiles in leaf and callus extracts of Phulai ( Acacia Modesta Wall). BMC Plant Biol 24:93. https://doi.org/10.1186/s12870-024-04747-9 Hazrati S, Mousavi Z, Mollaei S et al (2025) Effect of Phenological Variation on the Phytochemical Composition and Antioxidant Activity of Different Organs of Capparis spinosa L. Horticulturae 11:702. https://doi.org/10.3390/horticulturae11060702 Hesami M, Jones AMP (2021) Modeling and optimizing callus growth and development in Cannabis sativa using random forest and support vector machine in combination with a genetic algorithm. Appl Microbiol Biotechnol 105:5201–5212. https://doi.org/10.1007/s00253-021-11375-y Hesami M, Jones AMP (2020) Application of artificial intelligence models and optimization algorithms in plant cell and tissue culture. Appl Microbiol Biotechnol 104:9449–9485. https://doi.org/10.1007/s00253-020-10888-2 Higazy AE, El-Mahrouk ME, El-Banna AN et al (2023) Production of Black Cumin via Somatic Embryogenesis, Chemical Profile of Active Compounds in Callus Cultures and Somatic Embryos at Different Auxin Supplementations. Agronomy 13:2633. https://doi.org/10.3390/agronomy13102633 Ikeuchi M, Sugimoto K, Iwase A (2013) Plant callus: mechanisms of induction and repression. Plant Cell 25:3159–3173 Jeevan Ram PS, Singh S, Ali SA et al (2025) Machine Learning-Driven Screening of Promising Mutants in Nilgirianthus ciliatus Callus Cultures for Enhanced Phytochemicals Production. Physiol Plant 177:e70536. https://doi.org/10.1111/ppl.70536 Karous O, Ben Haj Jilani I, Ghrabi-Gammar Z (2021) Ethnobotanical study on plant used by semi-nomad descendants’ community in Ouled Dabbeb—Southern. Tunisia Plants 10:642 Katırcı R, Aasim M, Deveci G, Mustafa Z (2024) Comparing quantum machine learning and classical machine learning for in vitro regeneration of cowpea ( Vigna unguiculata ). Plant Cell Tiss Organ Cult 159:32. https://doi.org/10.1007/s11240-024-02880-9 Kaushik P, Rani M, Khurana N et al (2025) Revolutionizing Plant Tissue Culture: Harnessing Artificial Intelligence for Precision Propagation and Optimization. NPJ 15:e040624230642. https://doi.org/10.2174/0122103155302871240527094915 Kianersi F, Abdollahi MR, Mirzaie-asl A et al (2020) Biosynthesis of rutin changes in Capparis spinosa due to altered expression of its pathway genes under elicitors’ supplementation. Plant Cell Tiss Organ Cult 141:619–631 Kirkan B, Ceylan O, SARIKÜRKCÜ C, Bektas T (2021) Phenolic profile, antioxidant and enzyme inhibitory activity of the ethyl acetate, methanol and water extracts of Capparis spinosa L. Int J Secondary Metabolite 8:337–351 Koçak M, Yılmaz MC, Kuzğun C et al (2025) Synthetic seed production in Crataegus monogyna L. and prediction of regeneration of synthetic seeds with machine learning algorithms. Plant Cell Tiss Organ Cult 161. https://doi.org/10.1007/s11240-025-03049-8 Koufan M, Belkoura I, Mazri MA (2022) In Vitro Propagation of Caper ( Capparis spinosa L.): A Review. Horticulturae 8:737. https://doi.org/10.3390/horticulturae8080737 Labbafi M, Mehrafarin A, Badi H et al (2018) Improve germination of caper ( Capparis spinosa L.) seeds by different induction treatments of seed dormancy breaking. Trakia J Sci 16:71–74 Lee KY, Nam D-H, Jeon Y et al (2024) Exploring the Production of Secondary Metabolites from a Halophyte Tetragonia tetragonoides through Callus Culture. Horticulturae 10:244. https://doi.org/10.3390/horticulturae10030244 Liu W, He Y, Xiang J et al (2011) The physiological response of suspension cell of Capparis spinosa L. to drought stress. J Med Plants Res 5:5899–5906 Lundberg SM, Erion G, Chen H et al (2020) From Local Explanations to Global Understanding with Explainable AI for Trees. Nat Mach Intell 2:56–67. https://doi.org/10.1038/s42256-019-0138-9 Lundberg SM, Lee S-I (2017) A Unified Approach to Interpreting Model Predictions Ma Q, Grones P, Robert S (2018) Auxin signaling: a big question to be addressed by small molecules. J Exp Bot 69:313–328. https://doi.org/10.1093/jxb/erx375 Mahood HE, Sarropoulou V, Tzatzani T-T et al (2024) Biomass and polyphenolic enhancement in date palm ( Phoenix dactylifera L.) callus cultures through Fusarium oxysporum elicitation. Plant Cell Tiss Organ Cult 159:76. https://doi.org/10.1007/s11240-024-02934-y Moghadamnia Y, Kani SNM, Ghasemi-Kasman M et al (2019) The anti-cancer effects of capparis spinosa hydroalcoholic extract. Avicenna J med biotechnol 11:43 Mohaddab M, El Goumi Y, Gallo M et al (2022) Biotechnology and In Vitro Culture as an Alternative System for Secondary Metabolite Production. Molecules 27:8093. https://doi.org/10.3390/molecules27228093 Mohaddab M, Genva M, Malika F et al (2024) Capparis spinosa : A Rich Source of Phenolic Compounds-A Comprehensive Review of Its Phytochemistry, Health Benefits, and Biotechnological Applications. Biocatal Agric Biotechnol 103409 Movafeghi A, HABIBI G, ALI AM (2008) Plant regeneration of Capparis spinosa L. using hypocotyl explants Munasinghe SP, Somaratne S, Weerakoon SR, Ranasinghe C (2020) Prediction of chemical composition for callus production in Gyrinops walla Gaetner through machine learning. Inf Process Agric 7:511–522 Murashige T, Skoog F (1962) A Revised Medium for Rapid Growth and Bio Assays with Tobacco Tissue Cultures. Physiol Plant 15:473–497. https://doi.org/10.1111/j.1399-3054.1962.tb08052.x Noordijk B, Garcia Gomez ML, ten Tusscher KHWJ et al (2024) The rise of scientific machine learning: a perspective on combining mechanistic modelling with machine learning for systems biology. Front Syst Biol 4. https://doi.org/10.3389/fsysb.2024.1407994 Nowruzian A, Aalami A (2023) Effect of different treatments on dormancy breaking for seed germination enhancement and metabolite analysis of Capparis spinosa L. J Med plants By-Products 12:387–396 Özcan E, Atar HH, Ali SA, Aasim M (2023) Artificial neural network and decision tree–based models for prediction and validation of in vitro organogenesis of two hydrophytes— Hemianthus callitrichoides and Riccia fluitans . In Vitro Cell Dev Biol - Plant 59:547–562. https://doi.org/10.1007/s11627-023-10367-z Ozyigit II, Dogan I, Hocaoglu-Ozyigit A et al (2023) Production of secondary metabolites using tissue culture-based biotechnological applications. Front Plant Sci 14:1132555 Pasternak TP, Steinmacher D (2024) Plant Growth Regulation in Cell and Tissue Culture. Vitro Plants 13:327. https://doi.org/10.3390/plants13020327 Peng S, Rajjou L (2024) Advancing plant biology through deep learning-powered natural language processing. Plant Cell Rep 43:208 Pérez-Mejía N, Villarreal ML, Sánchez-Carranza JN et al (2024) Phytochemical Profiles and Cytotoxic Activity of Bursera fagaroides (Kunth) Engl. Leaves and Its Callus Culture. Plants 13:1622. https://doi.org/10.3390/plants13121622 Petch J, Di S, Nelson W (2022) Opening the Black Box: The Promise and Limitations of Explainable Machine Learning in Cardiology. Cana J Cardiol 38:204–213. https://doi.org/10.1016/j.cjca.2021.09.004 Rahimi VB, Rajabian A, Rajabi H et al (2020) The effects of hydro-ethanolic extract of Capparis spinosa ( C. spinosa ) on lipopolysaccharide (LPS)-induced inflammation and cognitive impairment: Evidence from in vivo and in vitro studies. J Ethnopharmacol 256:112706 Ranade R, Joshi N, Kudale S (2023) Comparative secondary metabolite expression in callus cultures and mother plant in Barleria prionitis L. Plant Cell Tiss Organ Cult 155:653–663. https://doi.org/10.1007/s11240-023-02585-5 Sadat-Hosseini M, Arab MM, Soltani M et al (2022) Predictive modeling of Persian walnut ( Juglans regia L.) in vitro proliferation media using machine learning approaches: a comparative study of ANN, KNN and GEP models. Plant Methods 18:48. https://doi.org/10.1186/s13007-022-00871-5 Sagharyan M, Ganjeali A, Cheniany M, Kouhi SMM (2020) Optimization of callus induction with enhancing production of phenolic compounds production and antioxidants activity in callus cultures of Nepeta binaloudensis Jamzad ( Lamiaceae ). Iran J Biotechnol 18:e2621 Sarabandi M, Singh RK, Kalantari S et al (2024) Unveiling machine learning’s impact on in vitro callogenesis optimization in Crocus sativus L. S Afr J Bot 168:1–8 Sarmah P, Borah T, Mudoi KD et al (2025) In vitro micropropagation of the critically endangered Aquilaria malaccensis Lam. and comparing the effect of PGRs’ through machine learning models. Plant Cell Tiss Organ Cult 162:6. https://doi.org/10.1007/s11240-025-03115-1 Slimani C, El Goumi Y, Rais C et al (2021) Optimization of callogenesis/caulogenesis induction protocol in saffron plant ( Crocus sativus L.) using response surface methodology. Biointerface Res Appl Chem 12:4731–4746 Sobhy FS, Ahmed S, Abdelrahim E et al (2025) Chemo Preventive Potential of Rutin-Rich Extract Obtained from Capparis spinosa Culture against HepG2 Cancer Cell Line Activity. Egypt J Chem 68:855–866 Sozzi GO, Peter KV, Babu KN, Divakaran M (2012) Capers and caperberries. Handbook of herbs and spices. Elsevier, pp 193–224 Sozzi GO, Vicente AR (2006) Capers and caperberries. Handbook of Herbs and Spices. Elsevier, pp 230–256 Svolacchia N, Sabatini S (2023) Cytokinins. Curr Biol 33:R10–R13. https://doi.org/10.1016/j.cub.2022.11.022 Tan WH, Tong CY, Chua MX, Derek CJC (2024) Modelling and optimizing secondary metabolites production in Spirodela polyrhiza using machine learning. J Clean Prod 478:143986. https://doi.org/10.1016/j.jclepro.2024.143986 Teoh SC, Subramaniam S, Chew BL (2023) The effects of 2, 4-dichlorophenoxyacetic acid on the induction of callus from cotyledon and hypocotyl explants of butterfly pea ( Clitoria ternatea ). Malays Appl Biol 52:61–72 Wang YT, Gan L, Liu W et al (2007) Research on the callus inducing and the cell growth and metabolism characteristics of Capparis spinosa L. College of Life Sci and Technolo, Huazhong Univ of Sci and Technolo. Wuhan China 7:1779–1783 Wijerathna-Yapa A, Hiti-Bandaralage J, Pathirana R (2025) Harnessing metabolites from plant cell tissue and organ culture for sustainable biotechnology. Plant Cell Tiss Organ Cult 162:55. https://doi.org/10.1007/s11240-025-03180-6 Yin Y, He Y, Liu W et al (2014) The durative use of suspension cells and callus for volatile oil by comparative with seeds and fruits in Capparis spinosa L. PLoS ONE 9:e113668 Yousefi E, Abedi M, Aghajanzadeh TA, Moreno DA (2025) Caper bush ( Capparis spinosa L.) bioactive compounds and antioxidant capacity as affected by adaptation to harsh soils. Sci Rep 15:11893. https://doi.org/10.1038/s41598-025-97298-4 Zarbakhsh S, Shahsavar AR, Soltani M (2024) Optimizing PGRs for in vitro shoot proliferation of pomegranate with bayesian-tuned ensemble stacking regression and NSGA-II: a comparative evaluation of machine learning models. Plant Methods 20:1–24. https://doi.org/10.1186/s13007-024-01211-5 Zhang H, Ma ZF (2018) Phytochemical and pharmacological properties of Capparis spinosa as a medicinal plant. Nutrients 10:116 Supplementary Files Supplemetarydata.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 11 Feb, 2026 Reviewers invited by journal 21 Jan, 2026 Editor assigned by journal 21 Jan, 2026 First submitted to journal 19 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8619922","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":578306144,"identity":"470a216b-1c5c-4b79-a988-80847b8310bf","order_by":0,"name":"Marouane Mohaddab","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3ElEQVRIiWNgGAWjYHACNiBiZjBgb2A8kADiSxCthecAA6laJBIYDjAQo0V32uFjD36UWcubS74xOPCgolaOQbr5AF4tZrfT0g17zqUb7pydY3Ag4cxxYwaZYwkEtOSYSfC2HWbccBuoJbHtWGKDRI4BQS2Sf9sO22+4eQao5d+x+gaJ/A8EtUgDbUnccIMHqKWhJoFBIgevDpBf0qRlzqUn7+xJKziQcOyAYZvMMUIOSz4m+abM2nY7++GND3/U1MnzSzc/wG8NGjgMiibSQB2pGkbBKBgFo2AEAACtIU59JfRShgAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-7791-368X","institution":"Universite de Liege Gembloux Agro-Bio Tech","correspondingAuthor":true,"prefix":"","firstName":"Marouane","middleName":"","lastName":"Mohaddab","suffix":""},{"id":578306145,"identity":"b7afc3f5-ac4c-4207-aefb-6cff2daf0fb7","order_by":1,"name":"Younes EL Goumi","email":"","orcid":"","institution":"Sultan Moulay Slimane University Polydisciplinary Faculty of Beni Mellal: Universite Sultan Moulay Slimane Faculte Polydisciplinaire de Beni Mellal","correspondingAuthor":false,"prefix":"","firstName":"Younes","middleName":"EL","lastName":"Goumi","suffix":""},{"id":578306147,"identity":"4683698b-1a5f-490d-a02f-c2c8a55cac59","order_by":2,"name":"Mohammed Elakrouch","email":"","orcid":"","institution":"Universite de Liege Gembloux Agro-Bio Tech","correspondingAuthor":false,"prefix":"","firstName":"Mohammed","middleName":"","lastName":"Elakrouch","suffix":""},{"id":578306148,"identity":"3f6cd9ef-2d17-4c09-baf2-99cd3dd70dff","order_by":3,"name":"Soufiane Hasni","email":"","orcid":"","institution":"Abdelmalek Essaadi University Faculty of Science and Technology of Tangier: Universite Abdelmalek Essaadi Faculte des Sciences et Techniques de Tanger","correspondingAuthor":false,"prefix":"","firstName":"Soufiane","middleName":"","lastName":"Hasni","suffix":""},{"id":578306153,"identity":"db2fa12a-b5e3-4b07-ad6a-b58ad044f547","order_by":4,"name":"Clément Burgeon","email":"","orcid":"","institution":"Universite de Liege Gembloux Agro-Bio Tech","correspondingAuthor":false,"prefix":"","firstName":"Clément","middleName":"","lastName":"Burgeon","suffix":""},{"id":578306156,"identity":"21f6de02-8ac7-42db-a72d-e3c9b24ef989","order_by":5,"name":"Manon Genva","email":"","orcid":"","institution":"Universite de Liege Gembloux Agro-Bio Tech","correspondingAuthor":false,"prefix":"","firstName":"Manon","middleName":"","lastName":"Genva","suffix":""},{"id":578306158,"identity":"ef1daf44-9958-4e9e-a5e3-67d317c3f9a6","order_by":6,"name":"Malika Fakiri","email":"","orcid":"","institution":"Universite Hassan 1er Faculte des Sciences et Techniques Settat","correspondingAuthor":false,"prefix":"","firstName":"Malika","middleName":"","lastName":"Fakiri","suffix":""},{"id":578306159,"identity":"3bff8056-47e3-4833-a016-a5b4b7f7af60","order_by":7,"name":"Marie-Laure Fauconnier","email":"","orcid":"","institution":"Universite de Liege Gembloux Agro-Bio Tech","correspondingAuthor":false,"prefix":"","firstName":"Marie-Laure","middleName":"","lastName":"Fauconnier","suffix":""}],"badges":[],"createdAt":"2026-01-16 14:20:55","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8619922/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8619922/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100980047,"identity":"d6d63c46-b97c-47e5-a010-51e91258c153","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":164079,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/db78f319b910594e5505c590.docx"},{"id":101203238,"identity":"db2e449f-af74-4b8d-bcfa-3b3d505bab0b","added_by":"auto","created_at":"2026-01-27 09:39:09","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":183111,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f1baad4407c036b7a9520a35.docx"},{"id":101296940,"identity":"8ac0f49b-4187-405b-9865-a85dcc9362d0","added_by":"auto","created_at":"2026-01-28 09:23:30","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63106,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.3.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/65e11c1f2a98adea79a4a368.docx"},{"id":100980051,"identity":"0bb65495-083b-4816-9d36-6e514bdbca10","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97152,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.4.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/d5cf3a00443a4b84c25d3d68.docx"},{"id":100980042,"identity":"febdec17-e304-4e38-8ad0-568c47b9961d","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":57416,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.5.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/97efbacfd7285ae82b3e8953.docx"},{"id":101203498,"identity":"e819b6bc-b246-4c50-af81-e70761a65971","added_by":"auto","created_at":"2026-01-27 09:39:54","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51107,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.6.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/e7ecef0307f4282a1cd5caa9.docx"},{"id":101203855,"identity":"f43b8f74-3072-4071-bc91-d5da2d08952e","added_by":"auto","created_at":"2026-01-27 09:40:48","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56396,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.7.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/d9c330369cce9621e00f01b1.docx"},{"id":100980041,"identity":"4da199ba-0e65-4266-9b02-5a6b364d1558","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13733,"visible":true,"origin":"","legend":"","description":"","filename":"pctoPCTOD2600033.xml","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/22889944b77fac06b21fd537.xml"},{"id":100980044,"identity":"934047a3-e8f1-4137-8f1c-e2f6953b3452","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1215,"visible":true,"origin":"","legend":"","description":"","filename":"PCTOD260003320916.go.xml","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f8b4160ca80251a12fadf93c.xml"},{"id":100980045,"identity":"406fcc65-f731-4e79-8821-9c1dda143828","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":906,"visible":true,"origin":"","legend":"","description":"","filename":"PCTOD2600033Import.xml","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/88524b7e5bbcc0c4cf5e84b6.xml"},{"id":101204057,"identity":"a0dcc13a-3404-4992-8bed-a5cfd17368e5","added_by":"auto","created_at":"2026-01-27 09:41:28","extension":"xml","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":173266,"visible":true,"origin":"","legend":"","description":"","filename":"PCTOD26000330enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/7b03f00bbe96696a7948637d.xml"},{"id":101203321,"identity":"dca5fb62-892d-44c1-ad5c-311dba31e705","added_by":"auto","created_at":"2026-01-27 09:39:23","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149290,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/1675f939fb73cb3dc9b12f1c.png"},{"id":100980049,"identity":"51d4bb54-ec4b-4ded-b9c7-0251a63d66a8","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48399,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/a754c0331ff6f6c6839b8db4.png"},{"id":100980056,"identity":"1fb22b70-3408-48f0-b10e-42cd26f4c220","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"jpeg","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82471,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f441a103e89545b5ed435b98.jpeg"},{"id":100980046,"identity":"85181a0e-e99f-4027-b5f5-0051802d392d","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":42761,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/7e0906a4b911b670531fd37a.png"},{"id":101203384,"identity":"a0c6230f-46f5-4745-b671-6cb5c41a2a3c","added_by":"auto","created_at":"2026-01-27 09:39:31","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36503,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/ab25acb1cdae59262165002d.png"},{"id":100980058,"identity":"3355d5b1-3491-4a08-a5e6-953edb5b8547","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41791,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/42ddedf3432388152a66c9cb.png"},{"id":100980052,"identity":"20355b66-1b00-48bf-a455-c2059fdc2391","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168404,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/9fe8af8afd998d42ac92a27a.png"},{"id":101203062,"identity":"5845e102-08cb-4d7b-94f4-8ee35d4fe177","added_by":"auto","created_at":"2026-01-27 09:38:41","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48399,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f45a39e650d3de7cf783abe9.png"},{"id":100980053,"identity":"160118e4-1de9-4fc8-ab49-7f1b2b713c39","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"jpeg","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82471,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/8984831ad41eb1910fecd552.jpeg"},{"id":101203839,"identity":"06113333-5126-4e4a-a427-500c2696296b","added_by":"auto","created_at":"2026-01-27 09:40:47","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":42761,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f96acda691f11d4a396ffc23.png"},{"id":100980060,"identity":"35205d26-22de-4bbe-abda-b4e6169c1e38","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36503,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/4e4507058980264550a1f746.png"},{"id":101203401,"identity":"31f587fd-ee40-4259-9355-326b109d53ea","added_by":"auto","created_at":"2026-01-27 09:39:33","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41791,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/87868fc38662011423ff627e.png"},{"id":100980066,"identity":"3e50f904-a368-40ca-a1b5-ed4fb981272a","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149290,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/9157c8385e4ae841c78bd604.png"},{"id":101203453,"identity":"f24bcacd-f11d-4fb5-924f-6387ce2c4bb7","added_by":"auto","created_at":"2026-01-27 09:39:47","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168404,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/ea9776b1de581e2ebea1e738.png"},{"id":100980067,"identity":"1f769a09-5c79-4ce3-bea7-7c4d7ba83ff7","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40747,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f141393dd1a9e8230183eda3.png"},{"id":101202964,"identity":"495f8c49-42a0-443d-9b46-e180f35b0698","added_by":"auto","created_at":"2026-01-27 09:38:18","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12136,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/5b2edeecc1d37e6ae3605637.png"},{"id":100980069,"identity":"63b6c5c5-eb83-409f-9b73-4038347459b8","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21571,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/ef53437ad05bdd59942f18cd.png"},{"id":101203791,"identity":"7cc7e42d-1249-4266-bd4b-75ade8261ad4","added_by":"auto","created_at":"2026-01-27 09:40:40","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11992,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/f6b9c95502210ff929c68b1a.png"},{"id":100980059,"identity":"d8d3e731-d824-44d3-9d5b-b05d575ef767","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11658,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/d9afc30498ecc4de002ff155.png"},{"id":100980057,"identity":"301ccf0a-7883-4a18-8a56-59fa8d646960","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11131,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/40cda01b329deb054ee40dea.png"},{"id":100980072,"identity":"d411f201-9ec4-44b1-b8b9-b923a8be5064","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43869,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/01b139c01d47f101533db0ba.png"},{"id":100980073,"identity":"ebe1311a-03a7-48a6-9571-b07ce133d939","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12136,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/045cabb71c06b6f397d5de63.png"},{"id":100980071,"identity":"5f197ee6-f753-451b-a680-2dd219c081cc","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21571,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/ae26472222157afd8a5e1986.png"},{"id":101203071,"identity":"7bf407bf-eb2e-4729-9da9-9da009a96a3d","added_by":"auto","created_at":"2026-01-27 09:38:42","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11992,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/915b05527532a627040013cf.png"},{"id":100980062,"identity":"3e102d47-26f2-47ea-9d86-c28b6b14ecc9","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11658,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/ae4e4b0261a9e61319063999.png"},{"id":101203437,"identity":"299eaa5d-ad79-4b15-8184-eacd8ab3c14c","added_by":"auto","created_at":"2026-01-27 09:39:40","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11131,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/8bd3e136032a05893c7a5733.png"},{"id":100980063,"identity":"4cdc8243-b243-4fa7-a4d3-6fa646c44069","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40747,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/2f0c6a7f3d57fe06c14627fa.png"},{"id":100980068,"identity":"66c72678-fb1c-4238-9e68-23c818ece5a4","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"png","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43869,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/419feb86a4c1d1a626dda095.png"},{"id":100980074,"identity":"3acc54cc-bb6e-47cb-b5e0-85da3d4fc8b0","added_by":"auto","created_at":"2026-01-23 12:03:40","extension":"xml","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":171382,"visible":true,"origin":"","legend":"","description":"","filename":"PCTOD26000330structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/31822a4c68239dd5bd3f0e58.xml"},{"id":101203178,"identity":"f21ca57b-bbe1-4f1d-80e0-0cc3afa302e0","added_by":"auto","created_at":"2026-01-27 09:38:59","extension":"html","order_by":42,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":183889,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/99faec8774fce097894f04d8.html"},{"id":100980036,"identity":"e261430e-1b49-465a-9801-2d1cd36dae6f","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1077237,"visible":true,"origin":"","legend":"\u003cp\u003eA schematic representation of the stepwise computational approach used in this study. (A) Modeling and prediction of callus Fresh Weight Gain (FWG) (output) based on 6-Benzylaminopurine (BAP), Kinetin (KIN), Naphthalene Acetic Acid (NAA), and 2,4-Dichlorophenoxyacetic Acid (2,4-D) as inputs using four models. (B) SHAP value analysis of the best-selected model to identify and quantify the key model features. (C) Stacking of the best model and using it as input for the production and accumulation of rutin (Model 2).\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/c4a48ea558f67d14669a2509.png"},{"id":101203233,"identity":"c5da9b5b-84e0-4f5b-8f50-e3879040633d","added_by":"auto","created_at":"2026-01-27 09:39:08","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1144184,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plots presenting the values of real vs predictions of callus fresh weight gain for four models: Random Forest, Gradient Boosting, XGBoost, and second-degree Polynomial regression.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/c273c83408e7c0eb72c52977.png"},{"id":100980035,"identity":"9dc970f3-9968-4dbc-b1cc-2660d9323cef","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":315374,"visible":true,"origin":"","legend":"\u003cp\u003eGlobal importance (A) and directional effects (B) of hormones on predicted callus FWG using SHAP values in Random Forest models. 6-benzylaminopurine (BAP), kinetin (KIN), naphthaleneacetic acid (NAA), and 2,4-dichlorophenoxyacetic acid (2,4-D).\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/372bdac0f11b0eea9fd42eff.png"},{"id":100980034,"identity":"24a20447-6f29-4605-9540-b8c4b9def33a","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":82471,"visible":true,"origin":"","legend":"\u003cp\u003eReal vs top 5 predicted combinations for maximization of callus fresh weight gain using random forest model.\u003c/p\u003e","description":"","filename":"image4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/eb13974a37ef6dbd98d6ac63.jpeg"},{"id":101203280,"identity":"1cf05d2c-4c65-4d24-bc78-98aaac2e4127","added_by":"auto","created_at":"2026-01-27 09:39:15","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":378047,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of real (green) and predicted (orange) rutin under different hormonal combinations using the random forest model.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/2a38a6e0b7c4e74bf424b7ef.png"},{"id":100980037,"identity":"f91a8af0-49eb-4461-b533-cd61aca6ac54","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":57338,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plots presenting the values of real vs predictions of rutin obtained by the random forest model.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/b20a13aff3e7e1242dfc13ae.png"},{"id":101203470,"identity":"891a351a-5fd5-4f43-a259-b475be352715","added_by":"auto","created_at":"2026-01-27 09:39:50","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":194196,"visible":true,"origin":"","legend":"\u003cp\u003eReal vs predicted accumulation of rutin using random forest model.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/9cb61cf04324b238320cf64d.png"},{"id":101751169,"identity":"bcd5fab6-4153-4921-b88f-7f1c2f747263","added_by":"auto","created_at":"2026-02-03 10:17:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3892701,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/57b0acd7-c1ad-4cc0-adeb-026e1c12ab83.pdf"},{"id":100980039,"identity":"c932d24d-d8b3-4663-96a0-cebc0f79b4d5","added_by":"auto","created_at":"2026-01-23 12:03:39","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":22970,"visible":true,"origin":"","legend":"","description":"","filename":"Supplemetarydata.docx","url":"https://assets-eu.researchsquare.com/files/rs-8619922/v1/11fcf44c7e7a5eccd22e7546.docx"}],"financialInterests":"","formattedTitle":"Integrating Machine Learning and SHAP Analysis to Boost Callus Growth and Rutin Biosynthesis in Capparis spinosa L.","fulltext":[{"header":"Introduction","content":"\u003cp\u003e \u003cem\u003eCapparis spinosa\u003c/em\u003e L., an endemic species of arid Mediterranean regions, has attracted increasing scientific attention due to its medicinal, nutritional, and agroecological values. The species is extensively harvested and commercially exploited across several Mediterranean countries, including Morocco, Spain, Italy, Greece, and Turkey (Sozzi and Vicente \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Grimalt et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Hazrati et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). \u003cem\u003eC. spinosa\u003c/em\u003e is characterized by a rich phytochemical profile, encompassing glucosinolates, phenolic acids, alkaloids, and flavonoids, which are associated with a wide range of biological activities such as antioxidants, antidiabetic, anti-inflammatory, and anticancer effects (Moghadamnia et al. \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Esmaeilpour et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Rahimi et al. \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). These properties highlight its considerable potential for applications in the pharmaceutical, nutraceutical, and cosmetic industries.(Karous et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Kirkan et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Annaz et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Among these compounds, rutin stands out for its high economic and therapeutic value, making it a compound of particular interest (Fotiadou et al. \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; El Amri et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The biosynthesis of rutin involves branched metabolic pathways from phenylalanine and is mediated by phenylalanine ammonia-lyase (PAL) and other key enzymes (Kianersi et al. \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe exploitation of this plant supports the livelihoods of rural and marginalized communities while contributing to local economies (Sozzi et al. \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). However, commercial harvesting relies mainly on wild populations, increasing pressure on natural resources and threatening long-term sustainability (Zhang and Ma \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Due to its tolerance to water deficit and adverse soils, \u003cem\u003eC. spinosa\u003c/em\u003e also represents an ideal candidate for climate-resilient agriculture (Yousefi et al. \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Despite these advantages, traditional propagation methods, whether sexual or vegetative, face significant limitations, including low efficiency, slow growth, and the potential loss of desirable traits (Gristina et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Pretreatments such as scarification, sulfuric acid exposure, or hormone application partially improve germination and rooting but remain insufficient for rapid large-scale multiplication (Labbafi et al. \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Nowruzian and Aalami \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cem\u003eIn vitro\u003c/em\u003e culture techniques, particularly callus induction, offer an effective solution (Efferth \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). They enable the regeneration of uniform, pathogen-free plants and may constitute effective reservoirs for the biosynthesis of secondary metabolites, thereby contributing to the conservation of natural populations and meeting the growing demand for phytotherapeutic resources (Yin et al. \u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Mohaddab et al. \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Although research on \u003cem\u003eC. spinosa\u003c/em\u003e remains limited (Caglar et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Movafeghi et al. \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Al-Safadi and Elias \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Carra et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Fahmideh et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Gianguzzi et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Mohaddab et al. \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), callus tissue, being totipotent, serves as an ideal platform for targeted secondary metabolite production. Hagaggi et al. (2024) reported that callus extracts contain higher levels of phenolic and flavonoid compounds, a greater diversity of volatile metabolites, and superior antioxidant and antibacterial activities compared to leaf extracts. Thus, strict control of culture conditions enhances the biosynthesis of bioactive flavonoids, offering direct potential for industrial and pharmaceutical applications (Wijerathna-Yapa et al. \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCallus formation and proliferation depend on intrinsic factors, including explant type, age, and medium composition, as well as extrinsic factors such as light, temperature, and humidity (Pasternak and Steinmacher \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Plant growth regulators (PGRs), particularly auxins and cytokinins, are essential for morphogenesis, biomass accumulation, and secondary metabolite production (Ikeuchi et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Bravo-V\u0026aacute;zquez et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Commonly used phytohormones include 2,4-dichlorophenoxyacetic acid (2,4-D) and 1-naphthaleneacetic acid (NAA), applied alone or in combination with cytokinins such as 6-benzylaminopurine (BAP) or kinetin (KIN) (Koufan et al. \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Higazy et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Studies indicate that 2,4-D at 1.0 mg/L combined with BAP at 1.5 mg/L effectively induces callus from \u003cem\u003eC. spinosa\u003c/em\u003e leaf explants (Wang et al. \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Yin et al. \u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Strategic manipulation of PGRs during callogenesis can therefore enhance both callus regeneration and flavonoid production, particularly rutin (Sobhy et al. \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cem\u003eIn vitro\u003c/em\u003e plant development is a complex process governed by multiple interconnected variables (Ma et al. \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Svolacchia and Sabatini \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Studying these factors in isolation is insufficient, as their interactions strongly influence outcomes. Classical statistical approaches are limited when handling nonlinear and multidimensional datasets and often exhibit high error rates (Aasim et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Katırcı et al. \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Mechanistic models, although grounded in biological knowledge, require precise measurement of numerous parameters that are difficult to standardize experimentally (Noordijk et al. \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), reducing their predictive power and practical usefulness for rapid protocol optimization (Baker et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Artificial intelligence (AI), especially supervised Learning or Machine Learning (ML), offers a means to manage this complexity (Peng and Rajjou \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). ML applications to callogenesis and organogenesis represent an emerging and promising field (Hesami and Jones \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Algorithms such as Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost) have been successfully applied to predict callus induction, shoot regeneration, biomass yield, and secondary metabolite production in various species (Hesami and Jones \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Alcalde et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Eren et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Demirel et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Sarabandi et al. \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Bozkurt et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Tan et al. \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zarbakhsh et al. \u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ko\u0026ccedil;ak et al. \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Sarmah et al. \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Despite their predictive performance, these models often remain \u0026ldquo;black boxes,\u0026rdquo; where relationships between biological variables and observed responses are opaque (Petch et al. \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). SHapley Additive exPlanations (SHAP) values provide a solution to this opacity. Derived from game theory, they quantify the contribution of each variable to the final prediction by considering all possible factor combinations (Lundberg and Lee \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lundberg et al. \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). By transforming a black-box model into a \u0026ldquo;white-box\u0026rdquo; model, SHAP values allow identification of key variables and their interactions, offering biologically meaningful interpretation (Ekanayake et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAlthough interest in \u003cem\u003eC. spinosa\u003c/em\u003e is increasing, investigations into callus culture and flavonoid biosynthesis, particularly rutin, remain limited and fragmented. The majority of studies have focused on isolated factors rather than the complex, non-linear interactions among growth regulators, explant type, and environmental conditions (Dar et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Furthermore, while machine learning approaches have demonstrated the capacity to drive diverse tissue culture systems, their application to \u003cem\u003eC. spinosa\u003c/em\u003e, especially using interpretable methods such as SHAP, has yet to be explored, leaving critical variables governing flavonoid production largely unexplored.\u003c/p\u003e \u003cp\u003eThis study examined the effects of growth regulators, individually and in combination, on callus induction and proliferation, as well as on rutin biosynthesis. The collected data were evaluated using machine learning models, complemented by SHAP analyses, to quantify the relative influence of each regulator and identify the key factors modulating both callus development and rutin biosynthesis.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003eCollection of Plant Material\u003c/p\u003e \u003cp\u003eFresh leaves of \u003cem\u003eC. spinosa\u003c/em\u003e L. were collected in April 2022 from the Taounate region (Morocco; latitude: 34\u0026deg;26'31.187\" N, longitude: 4\u0026deg;42'11.97\" W, altitude: 435 m). The leaves were harvested on the same day they were processed for further experiments.\u003c/p\u003e \u003cp\u003eSurface Sterilization\u003c/p\u003e \u003cp\u003eHarvested leaves were surface sterilized prior to culture. Sterilization was performed following the protocol described by Slimani et al. (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), ensuring that explants were free from microbial contamination before placement on culture medium.\u003c/p\u003e \u003cp\u003eCulture Medium Preparation\u003c/p\u003e \u003cp\u003eCallus induction was carried out on Murashige and Skoog (MS) medium ( Murashige and Skoog \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e1962\u003c/span\u003e), supplemented with 0.8% agar, 3% sucrose, and MES buffer (M0255, Duchefa). Leaf explants, approximately 1 cm in diameter, were placed abaxial side down on 20 mL of medium in each Petri dish, with five explants per dish. A total of 1,000 explants were used in the experiment. The pH of the medium was adjusted to 5.7\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1 before autoclaving at 121\u0026deg;C for 20 minutes under 1 bar of pressure.\u003c/p\u003e \u003cp\u003ePlant Growth Regulator Combination Design\u003c/p\u003e \u003cp\u003eThe basal medium was supplemented with different concentrations of exogenous plant growth regulators (PGRs) to evaluate their effects on callus induction. The PGRs tested included 2,4-dichlorophenoxyacetic acid (2,4-D: 0.0\u0026ndash;2.0 mg/L), 6-benzylaminopurine (BAP: 0.0\u0026ndash;2.0 mg/L), 1-naphthaleneacetic acid (NAA: 0.0\u0026ndash;2.0 mg/L), and kinetin (KIN: 0.0\u0026ndash;2.0 mg/L). A full factorial design was applied with four PGR combinations: BAP/NAA, BAP/2,4-D, KIN/NAA, and KIN/2,4-D. For each combination, four concentrations (0.0\u0026ndash;1.5 mg/L) were applied per regulator, resulting in 16 unique concentration pairs per combination. Each PGR was also tested individually at 2.0 mg/L, for a total of 20 distinct treatments, each replicated three times.\u003c/p\u003e \u003cp\u003e \u003cem\u003eIn vitro\u003c/em\u003e culture Conditions\u003c/p\u003e \u003cp\u003eAll cultures were maintained in a phytotron under a 16-hour light/8-hour dark photoperiod at 24\u0026thinsp;\u0026plusmn;\u0026thinsp;2\u0026deg;C for 30 days. Illumination was provided by Sylvania fluorescent tubes delivering a light intensity of approximately 1600 W.\u003c/p\u003e \u003cp\u003eMeasurement of Callus Growth\u003c/p\u003e \u003cp\u003eAfter 30 days, callus Fresh Weight Gain (FWG) was measured by weighing callus immediately after removal from the medium. These data were then compiled into a dataset for machine Learning analysis.\u003c/p\u003e \u003cp\u003eFWG was calculated as: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{F}\\text{W}\\text{G}=\\:\\text{F}\\text{i}\\text{n}\\text{a}\\text{l}\\:\\text{f}\\text{r}\\text{e}\\text{s}\\text{h}\\:\\text{w}\\text{e}\\text{i}\\text{g}\\text{h}\\text{t}\\:-\\:\\text{I}\\text{n}\\text{i}\\text{t}\\text{i}\\text{a}\\text{l}\\:\\text{w}\\text{e}\\text{i}\\text{g}\\text{h}\\text{t}\\:\\text{o}\\text{f}\\:\\text{e}\\text{x}\\text{p}\\text{l}\\text{a}\\text{n}\\text{t}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eIdentification and semi-quantification of rutin in callus Cultures using LC\u0026ndash;MS/MS\u003c/p\u003e \u003cp\u003eApproximately 150 mg of bone ash, dried at 30\u0026deg;C to constant weight, was extracted according to the protocol of Belkessam et al. (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), with minor modifications, using 2 mL of methanol\u0026ndash;water (80:20, v/v) and 5 \u0026micro;L of flavone as an internal standard. The samples were sonicated for 20 minutes, centrifuged for 10 minutes, and then filtered through 0.22 \u0026micro;m filters before analysis by LC\u0026ndash;MS/MS. Rutin was qualitatively identified by LC-QTOF and semi-quantified by LC-TQ via the MRM transition 609 \u0026rarr; 300, in accordance with the original method. Acquisition and processing of the data were carried out with MassHunter\u0026reg; Workstation software (version 10.0, Agilent Technologies, Santa Clara, CA, USA).\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\text{M}\\text{o}\\text{l}\\text{e}\\text{c}\\text{u}\\text{l}\\text{e}\\:\\text{c}\\text{o}\\text{n}\\text{t}\\text{e}\\text{n}\\text{t}\\:\\left(\\text{A}.\\text{U}./mg\\:\\text{D}\\text{W}\\right)=\\frac{\\text{M}\\text{R}\\text{M}\\:\\text{p}\\text{e}\\text{a}\\text{k}\\:\\text{a}\\text{r}\\text{e}\\text{a}\\:\\text{o}\\text{f}\\:\\text{r}\\text{u}\\text{t}\\text{i}\\text{n}}{\\text{M}\\text{R}\\text{M}\\:\\text{f}\\text{l}\\text{a}\\text{v}\\text{o}\\text{n}\\text{e}\\:\\text{p}\\text{e}\\text{a}\\text{k}\\:\\text{a}\\text{r}\\text{e}\\text{a}\\:\\text{x}\\:\\text{S}\\text{a}\\text{m}\\text{p}\\text{l}\\text{e}\\:\\text{d}\\text{r}\\text{y}\\:\\text{w}\\text{e}\\text{i}\\text{g}\\text{h}\\text{t}\\:}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eA.U. = Arbitrary Units; DW\u0026thinsp;=\u0026thinsp;Dry Weight ; MRM\u0026thinsp;=\u0026thinsp;Multiple Reaction Monitoring.\u003c/p\u003e \u003cp\u003eModeling procedures\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe study methodology is presented in figure. 1. Callogenesis was induced using four PGR combinations: BAP/NAA, BAP/2,4-D, KIN/NAA, and KIN/2,4-D at varying concentrations to evaluate their effects on FWG and rutin accumulation. To ensure data quality, the data were normalized, outliers were removed, and the relationships between parameters were examined using a Pearson correlation test. The final dataset was partitioned into a training set (75%) and an independent test set (25%), while maintaining a balanced distribution of experimental conditions (Figure. 1 (A)).\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e1\u003c/span\u003eA schematic representation of the stepwise computational approach used in this study. (A) Modeling and prediction of callus Fresh Weight Gain (FWG) (output) based on 6-Benzylaminopurine (BAP), Kinetin (KIN), Naphthalene Acetic Acid (NAA), and 2,4-Dichlorophenoxyacetic Acid (2,4-D) as inputs using four models. (B) SHAP value analysis of the best-selected model to identify and quantify the key model features. (C) Stacking of the best model and using it as input for the production and accumulation of rutin (Model 2).\u003c/p\u003e \u003cp\u003eFour machine learning models, RF, GB, XGBoost, and second-degree polynomial regression (PL2), were evaluated for FWG prediction. Model performance was assessed using established statistical indicators: the coefficient of determination (R\u0026sup2;) and cross-validation (R\u0026sup2;CV), moreover root mean square error (RMSE). Low RMSE values indicate that the model\u0026rsquo;s predictions are reliable, with minimal error. R\u0026sup2; quantifies the proportion of the total variance in the observed data captured by the model, while R\u0026sup2;CV assesses its ability to generalize to new data. Both range from 0 to 1, with values closer to 1 reflecting strong agreement between predicted and observed values and good model stability. Overall, a model with high R\u0026sup2; and R\u0026sup2;CV combined with low RMSE can accurately and reliably reproduce the behavior of the studied system.\u003c/p\u003e \u003cp\u003eSHAP effectively serves as a sensitivity analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e1\u003c/span\u003e(B)), measuring the impact of each growth regulator on FWG while accounting for the combined effects of all other variables. This approach allows the identification of regulators with the greatest influence on callus growth and provides insight into the complex interactions among them.\u003c/p\u003e \u003cp\u003ePredictions from model 1 were then incorporated into a stacked model designed to predict rutin accumulation (model 2) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e1\u003c/span\u003e(C)). This model used both the initial concentrations of the growth regulators and the FWG predictions as input variables, directly linking callus growth to secondary metabolite production. To validate the efficacy and reliability of this optimized model, the predicted conditions for maximizing FWG and rutin accumulation were experimentally tested in triplicate. The results confirmed the accuracy of the predictions and demonstrated the practical utility of the model for optimizing callogenesis and secondary metabolite production.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eInfluence of plant growth regulators on \u003cem\u003eC. Spinosa\u003c/em\u003e callus Induction\u003c/p\u003e \u003cp\u003eThis study investigated the influence of phytohormones matrix at different concentrations (0, 0.5, 1 and 1.5 mg/L) and hormonal combinations (BAP/NAA), (BAP/2,4-D), (KIN/NAA), and (KIN/2,4-D) on the FWG of callus derived from \u003cem\u003eC. spinosa\u003c/em\u003e leaf explants after 30 days of culture (Table A). Notably, no callus formation occurred in the culture medium lacking PGRs, highlighting the essential role of exogenous hormones in inducing callogenesis in \u003cem\u003eC. spinosa\u003c/em\u003e leaf explants.\u003c/p\u003e \u003cp\u003eAmong the tested hormone combinations, 2,4-D and BAP showed an effect on callus induction and growth. The highest FWG (3639.5 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;208.0) was recorded with a combination of 1 mg/L BAP with 1 mg/L 2,4-D. Similarly high values were observed with 1 mg/L BAP\u0026thinsp;+\u0026thinsp;0.5 mg/L 2,4-D (3459.33 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;84.68). Additionally, 2,4-D applied alone also promoted substantial callus proliferation, with FWG values of 3438.5 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;493.5 and 3394 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;31 at 0.5 and 1 mg/L, respectively.\u003c/p\u003e \u003cp\u003eIn contrast, the (BAP/ NAA) and (KIN/ NAA) combinations produced significantly lower FWG, indicating less efficiency in promoting callus formation and proliferation. The maximum FWG in these groups was observed with 0.5 mg/L BAP combined with 1.5 NAA (1874.45 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;25.66) and with 0.5 mg/L KIN combined with 1 mg/L NAA (1350.5 mg\u0026thinsp;\u0026plusmn;\u0026thinsp;215.5). To further investigate these complex hormonal interactions and their effects on callus induction, artificial intelligence was employed to analyze the dataset.\u003c/p\u003e \u003cp\u003eModel performance in predicting callus induction\u003c/p\u003e \u003cp\u003eThe callogenesis of \u003cem\u003eC. spinosa\u003c/em\u003e was quantitatively modeled using a set of explanatory variables representing four exogenously applied PGRs: BAP, KN, NAA, and 2,4-D. The target variable for modeling was the FWG of callus tissue. To capture the complex relationships between hormonal treatments (inputs) and FWG (outputs), four ML models were applied and compared: RF, GB, XGBoost, and a traditional second-degree Polynomial Regression.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMachine learning model performance for predicting callus fresh weight gain.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eML Models\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e CV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRMSE\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e263.33\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e285.59\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGBoost\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e274.46\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePL2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e471.98\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eR\u0026sup2;: coefficient of determination, R\u0026sup2;CV: cross-validated R\u0026sup2;, RMSE: root mean square error, RF: Random Forest, GB: Gradient Boosting, XGBoost: Extreme Gradient Boosting, PL2: second-degree polynomial regression.\u003c/p\u003e \u003cp\u003eThe ensemble Learning models, RF, GB, and XGBoost, demonstrated superior and comparable performance (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). RF had the highest predictive accuracy with an R\u0026sup2; of 0.86, R\u0026sup2;CV of 0.89, and a low RMSE of 263.33, indicating precise predictions and strong generalization. XGBoost and GB also performed well, with R\u0026sup2; values of 0.85 and 0.83, R\u0026sup2;CV of 0.86 for both, and RMSE values of 274.46 and 285.59, respectively. These results highlight the ability of ensemble methods to capture complex, non-linear relationships in FWG data effectively. In contrast, the PL2 model exhibited substantially lower performance, with an R\u0026sup2; of 0.56, R\u0026sup2;CV of 0.63, and a markedly higher RMSE of 471.98. This indicates that simple polynomial relationships are insufficient to model the underlying biological complexity, resulting in larger prediction errors. Overall, the findings suggest that ensemble learning approaches, particularly RF and XGBoost, provide reliable and robust predictions for FWG.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e2\u003c/span\u003e Scatter plots presenting the values of real vs predictions of callus fresh weight gain for four models: Random Forest, Gradient Boosting, XGBoost, and second-degree Polynomial regression.\u003c/p\u003e \u003cp\u003eThe combined analysis of numerical and graphical results highlights the superiority of ensemble tree-based models (RF, GB, and XGBoost) for predicting callus FWG. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e2\u003c/span\u003e, these models exhibit regression slopes close to unity (RF: 0.98 x\u0026thinsp;+\u0026thinsp;42.04; GB: 0.94 x\u0026thinsp;+\u0026thinsp;83.13; XGBoost: 0.96 x\u0026thinsp;+\u0026thinsp;71.62) and a tight clustering of points around the ideal y\u0026thinsp;=\u0026thinsp;x line, indicating precise calibration and limited residual dispersion. These findings are consistent with the high coefficients of determination and low cross-validation errors (R\u0026sup2;CV) reported in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, confirming the robustness and generalizability of these models across the full FWG range. In contrast, PL2 shows a markedly lower slope (0.81 x\u0026thinsp;+\u0026thinsp;244.11) and substantial dispersion, reflecting systematic underestimation of high FWG values and poor ability to capture the non-linearities inherent in biological data. Overall, these results indicate that ensemble models, particularly Random Forest, represent the most accurate and reliable approach for predicting callus fresh weight gain, whereas polynomial regression remains unsuitable for modeling such complex relationships.\u003c/p\u003e \u003cp\u003e The graphs show the overall importance of the variables (Figure. 2A) reveal that 2,4-D has the highest absolute mean SHAP value (~\u0026thinsp;700), significantly surpassing those of BAP, KIN, and NAA, making it the most influential factor in predicting callus biomass accumulation. Analyzing the meaning of SHAP values (Figure. 3B), 2,4-D has a strongly positive effect on FWG (+\u0026thinsp;12.19). Comparatively, BAP also contributes positively, but more moderately (+\u0026thinsp;6.39). In contrast, KIN reduces predicted growth (-6.95), and NAA has a slightly negative impact (-2.01), highlighting its marginal role. Thus, these results confirm that 2,4-D is the key regulator of callus growth, BAP plays a secondary favorable role, while KIN may act as an inhibitor, and NAA remains of little influence.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e3\u003c/span\u003e Global importance (A) and directional effects (B) of hormones on predicted callus FWG using SHAP values in Random Forest models. 6-benzylaminopurine (BAP), kinetin (KIN), naphthaleneacetic acid (NAA), and 2,4-dichlorophenoxyacetic acid (2,4-D).\u003c/p\u003e \u003cp\u003eExperimental validation of callus fresh weight gain\u003c/p\u003e \u003cp\u003eLaboratory experiments were conducted to validate the five top phytohormone combinations predicted by the RF model and assess the reliability of the results. The validation experiment results (Figure. 4) demonstrated the reliability and accuracy of RF in predicting optimal combinations of plant growth regulators, enabling the achievement of maximal callus FWG. Results from the validation experiment provide robust evidence for the effectiveness of decision tree models, particularly Random Forest, as a powerful tool for optimizing callus induction. This approach also paves the way for suspension culture and large-scale production, while further enabling the optimization of micropropagation protocols for plants, such as vegetative propagation of \u003cem\u003eC. Spinosa\u003c/em\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCallus culture as a platform for rutin accumulation\u003c/p\u003e \u003cp\u003eHormonal regulation is a key factor in controlling the biosynthesis of secondary metabolites in \u003cem\u003eC. spinosa\u003c/em\u003e callus cultures. In this study, rutin, the main metabolite of this species, was first identified using UPLC-QTOF-MS/MS, while relative concentration was then semi-quantified using UPLC-TQ-MS/MS in MRM mode (609 \u0026rarr; 300). The BAP/2,4-D hormone combination was chosen for these experiments due to its favorable influence on FWG (Figure. 3), creating optimal conditions to assess rutin accumulation in callus tissues. The influence of hormonal treatments on rutin content was further analyzed using an RF model. Figure. 5 presents a comparison between experimentally determined and predicted rutin concentrations. The data indicate that rutin accumulation is strongly dependent on the relative concentrations of BAP and 2,4-D. Maximum rutin levels were observed under moderate BAP concentrations (0.5\u0026ndash;1.0 mg/L) combined with low 2,4-D concentrations (0.0\u0026ndash;1.0 mg/L), suggesting that these conditions favor activation of the flavonoid biosynthetic pathway. In contrast, higher concentrations of either BAP (\u0026ge;\u0026thinsp;1.5 mg/L) or 2,4-D (\u0026ge;\u0026thinsp;1.5 mg/L) markedly reduced rutin content, implying an inhibitory effect of excessive hormone levels on secondary metabolism. Overall, the RF model effectively captured these trends, with predicted values closely aligning with experimental observations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure. 6 further illustrates the predictive performance of the RF model. Observed versus predicted rutin concentrations cluster around the ideal 1:1 line, underscoring the model\u0026rsquo;s ability to faithfully reproduce the relationship between measured and estimated values. Regression analysis yielded a slope of 0.79 and an intercept of 0.25, with a high coefficient of determination (R\u0026sup2; = 0.821) and a highly significant \u003cem\u003ep\u003c/em\u003e-value, confirming the statistical robustness and reliability of the model. These results demonstrate that the model accurately captures overall trends, even within the complexity of a biological system. Ensemble tree-based algorithms are notable for their flexibility, capacity to model high-order nonlinear interactions, ability to handle heteroscedasticity, and robustness against overfitting. These qualities render the model a powerful tool for predicting and understanding the biosynthesis of secondary metabolites, such as rutin, in plant tissue cultures.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eExperimental validation of rutin accumulation in callus cultures\u003c/p\u003e \u003cp\u003eThe predictions of rutin accumulation obtained by the RF model, based on the BAP/2, 4-D combinations presented in section 3.3, show overall agreement with the experimental data, particularly for the condition BAP\u0026thinsp;=\u0026thinsp;2.0\u0026thinsp;+\u0026thinsp;2.4-D\u0026thinsp;=\u0026thinsp;2.0, highlighting the robustness of the model in anticipating metabolic responses to different hormonal treatments. The slight overestimation observed for certain combinations (BAP\u0026thinsp;=\u0026thinsp;1.5\u0026thinsp;+\u0026thinsp;2.4-D\u0026thinsp;=\u0026thinsp;2.0 and BAP\u0026thinsp;=\u0026thinsp;2.0\u0026thinsp;+\u0026thinsp;2.4-D\u0026thinsp;=\u0026thinsp;1.0) does not alter the overall consistency of the predictions. These results demonstrate that the model is a reliable tool for estimating rutin accumulation in \u003cem\u003eC. spinosa\u003c/em\u003e callus cultures and provide effective support for optimizing culture conditions to maximize secondary metabolite production.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eCallus induction is a central step in plant tissue culture, representing a cellular reprogramming process in which differentiated cells revert to an undifferentiated and proliferative state (Lee et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This process is primarily regulated by the exogenous application of PGRs, which provide the plasticity required for subsequent morphogenetic events, such as organogenesis and somatic embryogenesis (Pasternak and Steinmacher \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Callus also serves as a reservoir for secondary metabolite production, including rutin. Therefore, investigating the effects of hormones, cytokinins, and auxins on callus induction and rutin accumulation is crucial for effective utilization of this biological resource.\u003c/p\u003e \u003cp\u003eIn this study, we evaluated the effects of varying concentrations of BAP, KIN, NAA, and 2,4-D on FWG and rutin accumulation in \u003cem\u003eC. spinosa\u003c/em\u003e using RF, GB, XGBoost, and PL2. ML algorithms, including RF and XGBoost, have been studied in plant tissue culture, yet comparative analyses remain limited.\u003c/p\u003e \u003cp\u003eThese findings highlight that classic statistical methods, based on linear relationships or limited curvature, struggle to capture the complex interactions, threshold effects, and saturation phenomena characteristic of plant hormonal responses (Sadat-Hosseini et al. \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Modeling callus formation \u003cem\u003ein vitro\u003c/em\u003e, where multiple hormonal factors interact in a nonlinear and potentially synergistic manner, requires more flexible approaches (Fallah Ziarani et al. \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Machine learning models, particularly tree-based ensemble algorithms such as RF and XGBoost, have proved particularly well suited to this challenge due to their ability to model high-order interactions, handle heteroscedasticity, and remain robust to overfitting (Munasinghe et al. \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ali and Aasim \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kaushik et al. \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The demonstrated accuracy of these models highlights the value of integrating machine learning into plant tissue culture research, not only to improve predictive ability but also to improve our understanding of the complex mechanisms driving hormonal responses.\u003c/p\u003e \u003cp\u003eTo evaluate the relative contributions of individual growth regulators to callus development, we focused the SHAP analysis on the RF model, selected for its superior performance in terms of predictive accuracy. This approach allows for a quantitative assessment of the effect of each factor (BAP, KIN, NAA, and 2,4-D) on FWG predictions for \u003cem\u003eC. spinosa\u003c/em\u003e callus tissue, while taking into account potential interactions between variables.\u003c/p\u003e \u003cp\u003eRecent studies comparing machine learning models for modeling and predicting saffron (\u003cem\u003eCrocus sativus\u003c/em\u003e) callogenesis demonstrated that XGBoost and GB achieved superior accuracy (R\u0026sup2; \u0026gt; 0.95) compared to artificial neural networks and RF regression in modeling \u003cem\u003ein vitro\u003c/em\u003e culture systems (Sarabandi et al. \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In our study, we compare the predictive performance of RF, GB, XGBoost, and PL2 for FWG in \u003cem\u003eC. spinosa\u003c/em\u003e. Tree-based Learning models consistently outperformed polynomial regression, with RF achieving the highest predictive accuracy (R\u0026sup2; \u0026gt; 0.86) for callogenesis. Although this represents the first direct comparison of these models in \u003cem\u003eC. spinosa in vitro\u003c/em\u003e culture, other studies have similarly reported the superior accuracy of RF (\u0026Ouml;zcan et al. \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Bozkurt et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Sarmah et al. \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Overall, a model\u0026rsquo;s predictive capability is largely determined by the size and complexity of the training dataset (Hesami and Jones \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In this context, the RF model has consistently demonstrated a remarkable ability to accurately predict complex callus-related biological outcomes, achieving a high coefficient of determination (R\u0026sup2; = 0.97) for both callus growth parameters and secondary metabolite production in \u003cem\u003eNilgirianthus ciliatus\u003c/em\u003e (Jeevan Ram et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These results confirm RF as a robust and reliable tool for predicting callus growth and metabolite accumulation, providing confidence in its application for optimizing \u003cem\u003ein vitro\u003c/em\u003e culture conditions.\u003c/p\u003e \u003cp\u003eSensitivity analyses were applied to interpret the RF model and quantify the contribution of each variable to FWG predictions. SHAP values as an interpretative tool for identifying and quantifying variable importance have been demonstrated in multiple studies (Ekanayake et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Antonini et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). SHAP provides a mathematically rigorous, objective, and consistent method to decompose model predictions into contributions from each variable (Lundberg and Lee \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This approach enables precise local explanation of individual predictions while offering a global perspective on variable importance, making it robust and broadly applicable to machine learning interpretation (Lundberg et al. \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Using SHAP, key variables such as 2,4-D, NAA, sucrose concentration, and culture duration have been successfully identified (Sarabandi et al. \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In our study, SHAP analysis confirmed that 2,4-D was the most influential factor, KIN exhibited inhibitory effects, and BAP acted synergistically with 2,4-D to enhance callus formation. Identification of these factors is particularly relevant when secondary metabolite production is the primary objective of callus culture.\u003c/p\u003e \u003cp\u003eTo our knowledge, this study provides the first systematic assessment of the impact of various PGRs on both callus induction and rutin production in \u003cem\u003eC. spinosa\u003c/em\u003e. To directly link callus growth with metabolite production, we employed a stacked RF model, using callogenesis outputs as inputs to predict rutin accumulation. This approach effectively modeled the nonlinear relationship between cell proliferation and rutin biosynthesis, yielding high predictive performance (R\u0026sup2; \u0026gt; 0.82) and minimal deviation between observed and predicted values, confirming method robustness.\u003c/p\u003e \u003cp\u003eFor instance, Yin et al. (\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) reported that MS medium supplemented with 1.5 mg/L 2,4-D and 3.0 mg/L BAP induces callogenesis and that callus may serve as a source of volatile organic compounds. Similarly, Duran and Issah (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) demonstrated that BAP at 1 mg/L combined with 2 mg/L NAA produced a callogenesis rate of approximately 120 mg after 10 days. While these studies highlighted the impact of the strigolactone GR24 on callus production and phenolic content in \u003cem\u003eC. spinosa\u003c/em\u003e, our results showed that BAP\u0026thinsp;+\u0026thinsp;NAA achieved over 1000 mg FWG over 30 days, although it was less effective than BAP\u0026thinsp;+\u0026thinsp;2,4-D. This confirms that NAA is less favorable for callogenesis. This combination also resulted in substantial rutin accumulation, with a modest increase in the presence of 0.1 \u0026micro;M GR24, indicating that \u003cem\u003eC. spinosa\u003c/em\u003e callus can serve as a rutin reservoir (Duran and Issah \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDirect comparison between studies is limited due to differences in explant source, methodology, and objectives. In the present work, maximum callus formation was observed with 2,4-D, consistent with reports in \u003cem\u003eC. spinosa\u003c/em\u003e (Sobhy et al. \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and other species where auxin application significantly promotes callogenesis (Slimani et al. \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Teoh et al. \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Ranade et al. \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Mahood et al. \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Abdelazeez et al. \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRegarding rutin accumulation, low concentrations of BAP and 2,4-D (\u0026le;\u0026thinsp;1 mg/L) promoted significant accumulation, peaking at 0.5 mg/L for each hormone, paralleling biomass increases. Similar trends were reported by Kianersi et al. (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), with further enhancements achieved using elicitors such as methyl jasmonate. (Goda et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) also reported that 2,4-D-induced callus exhibited significantly higher total flavonoid content. These findings highlight the importance of optimizing hormone concentrations and combining them with elicitors for large-scale production of high-quality bioactive compounds (Sagharyan et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ozyigit et al. \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; P\u0026eacute;rez-Mej\u0026iacute;a et al. \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe reliability of RF predictions was validated experimentally. Predicted maximal FWG results closely matched observed values, demonstrating the model\u0026rsquo;s accuracy. For rutin accumulation, satisfactory performance was observed, particularly for BAP\u0026thinsp;+\u0026thinsp;2,4-D at 2 mg/L, with minor overestimations in some combinations. These results confirm RF as a robust and effective machine learning algorithm for modeling and optimizing callus culture in this study.\u003c/p\u003e \u003cp\u003eFuture studies using larger datasets and additional variables could further improve model performance. Investigations on \u003cem\u003eC. spinosa\u003c/em\u003e could explore 2,4-D with other PGRs or extend to cell suspension cultures and bioreactor-based production systems.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis research provides the first comprehensive evaluation of the combined effects of BAP, KIN, NAA, and 2,4-D on callogenesis and rutin biosynthesis in \u003cem\u003eC. spinosa\u003c/em\u003e using an integrated machine learning and explainable-AI framework. Ensemble tree-based models, particularly Random Forest, accurately predicted callus biomass and metabolite accumulation, outperforming classical statistical approaches. SHAP analysis revealed 2,4-D as the primary determinant of callus proliferation, while BAP acted synergistically to enhance growth, and KIN and NAA contributed weak or inhibitory effects. Experimental validation confirmed the reliability of the predicted optimal hormonal combinations, demonstrating strong agreement between predicted and observed FWG and rutin levels. Moderate concentrations of BAP and 2,4-D favored rutin accumulation, underscoring the importance of fine hormonal tuning for metabolic optimization.\u003c/p\u003e \u003cp\u003eOverall, this study highlights the relevance of ML-driven modeling for deciphering complex hormonal interactions and accelerating protocol optimization in plant tissue culture. The established models constitute a robust foundation for scaling up callus-based production systems, including suspension cultures and bioreactors, aimed at sustainable, high-yield biosynthesis of rutin and other valuable metabolites in \u003cem\u003eC. spinosa\u003c/em\u003e. Future investigations integrating larger datasets, molecular markers, and elicitation strategies will further strengthen predictive performance and advance the biotechnological exploitation of this species.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCRediT authorship contribution statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMarouane Mohaddab:\u0026nbsp;\u003c/strong\u003eWriting \u0026ndash; review \u0026amp; editing, Writing \u0026ndash; original draft, Conceptualization, Methodology,\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eData curation, Formal analysis. \u003cstrong\u003eYounes EL Goumi:\u0026nbsp;\u003c/strong\u003eSupervision, Writing \u0026ndash; review and editing, Methodology. \u003cstrong\u003eMohammed Elakrouch:\u0026nbsp;\u003c/strong\u003eFormal analysis, Software, Methodology \u003cstrong\u003eSoufiane Hasni:\u0026nbsp;\u003c/strong\u003eFormal analysis, Software, Data curation. \u003cstrong\u003eCl\u0026eacute;ment Burgeon:\u003c/strong\u003e Writing \u0026ndash; review \u0026amp; editing, Software, Data curation. \u003cstrong\u003eManon Genva:\u0026nbsp;\u003c/strong\u003eWriting \u0026ndash; review \u0026amp; editing, Software. \u003cstrong\u003eMalika Fakiri:\u003c/strong\u003e Writing \u0026ndash; review \u0026amp; editing, Validation, Project administration\u003cstrong\u003e\u0026nbsp;Marie-Laure Fauconnier:\u0026nbsp;\u003c/strong\u003eWriting \u0026ndash; review \u0026amp; editing, Validation\u003cstrong\u003e,\u0026nbsp;\u003c/strong\u003eProject administration\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declaration \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that there is no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAasim M, Ali SA, Altaf MT et al (2023) Artificial neural network and decision tree facilitated prediction and validation of cytokinin-auxin induced \u003cem\u003ein vitro\u003c/em\u003e organogenesis of sorghum (\u003cem\u003eSorghum bicolor\u003c/em\u003e L). Plant Cell Tiss Organ Cult 153:611\u0026ndash;624. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-023-02498-3\u003c/span\u003e\u003cspan address=\"10.1007/s11240-023-02498-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbdelazeez WMA, Aboueldis GR, Suliman AA, Mohammed DM (2025) Production of secondary metabolites in callus cultures of \u003cem\u003eScutellaria baicalensis\u003c/em\u003e L. and assessment of their anti-inflammatory and antioxidant efficacy in ulcerative colitis rats. Plant Cell Tiss Organ Cult 160:80. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-025-02996-6\u003c/span\u003e\u003cspan address=\"10.1007/s11240-025-02996-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlcalde MA, M\u0026uuml;ller M, Munn\u0026eacute;-Bosch S et al (2022) Using machine learning to link the influence of transferred \u003cem\u003eAgrobacterium rhizogenes\u003c/em\u003e genes to the hormone profile and morphological traits in \u003cem\u003eCentella asiatica\u003c/em\u003e hairy roots. Front Plant Sci 13:1001023\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli SA, Aasim M (2024) Response surface methodology and artificial intelligence modeling for \u003cem\u003ein vitro\u003c/em\u003e regeneration of Brazilian micro sword (\u003cem\u003eLilaeopsis brasiliensis\u003c/em\u003e). Plant Cell Tiss Organ Cult 157:10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-024-02734-4\u003c/span\u003e\u003cspan address=\"10.1007/s11240-024-02734-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAl-Safadi B, Elias R (2011) Improvement of caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) propagation using \u003cem\u003ein vitro\u003c/em\u003e culture and gamma irradiation. Scientia Horti 127:290\u0026ndash;297\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnaz H, Sane Y, Bitchagno GTM et al (2022) Caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.): an updated review on its phytochemistry, nutritional value, traditional uses, and therapeutic potential. Front Pharmacol. 2022; 13: 1\u0026ndash;22\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAntonini AS, Tanzola J, Asiain L et al (2024) Machine learning model interpretability using SHAP values: Application to Igneous Rock Classification task. Appl Comput Geosci 23:100178. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.acags.2024.100178\u003c/span\u003e\u003cspan address=\"10.1016/j.acags.2024.100178\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaker RE, Pe\u0026ntilde;a J-M, Jayamohan J, J\u0026eacute;rusalem A (2018) Mechanistic models versus machine learning, a fight worth fighting for the biological community? Biol Lett 14:20170660. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1098/rsbl.2017.0660\u003c/span\u003e\u003cspan address=\"10.1098/rsbl.2017.0660\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBelkessam M, Genva M, Kouki A et al (2025) Phytochemical Profile and Cosmeceutical Potential of Leaf Extracts of Two Species of the \u003cem\u003eAnacardiaceae\u003c/em\u003e Family from the Mediterranean Scrubland: \u003cem\u003ePistacia lentiscus\u003c/em\u003e L. and \u003cem\u003ePistacia atlantica\u003c/em\u003e Desf. Processes 13:3712. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/pr13113712\u003c/span\u003e\u003cspan address=\"10.3390/pr13113712\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBozkurt T, İnan S, D\u0026uuml;ndar İ et al (2024) Optimizing the \u003cem\u003eIn Vitro\u003c/em\u003e Propagation of Tea Plants: A Comparative Analysis of Machine Learning Models. Horticulturae 10:721. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/horticulturae10070721\u003c/span\u003e\u003cspan address=\"10.3390/horticulturae10070721\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBravo-V\u0026aacute;zquez LA, Angulo\u0026ndash;Bejarano PI, Bandyopadhyay A et al (2023) Regulatory roles of noncoding RNAs in callus induction and plant cell dedifferentiation. Plant Cell Rep 42:689\u0026ndash;705. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00299-023-02992-0\u003c/span\u003e\u003cspan address=\"10.1007/s00299-023-02992-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCaglar G, Caglar S, Ergin O, Yarim M (2005) The influence of growth regulators on shoot proliferation and rooting of \u003cem\u003ein vitro\u003c/em\u003e propagated caper. J Environ Biol 26:479\u0026ndash;485\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarra A, Del Signore MB, Sottile F et al (2012) Potential use of new diphenylurea derivatives in micropropagation of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. Plant Growth Regul 66:229\u0026ndash;237. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10725-011-9645-3\u003c/span\u003e\u003cspan address=\"10.1007/s10725-011-9645-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDar SA, Nawchoo IA, Tyub S, Kamili AN (2021) Effect of plant growth regulators on \u003cem\u003ein vitro\u003c/em\u003e induction and maintenance of callus from leaf and root explants of \u003cem\u003eAtropa acuminata\u003c/em\u003e Royle ex Lindl. Biotechnol Rep 32:e00688\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemirel F, Uğur R, Popescu GC et al (2023) Usage of Machine Learning Algorithms for Establishing an Effective Protocol for the \u003cem\u003eIn Vitro\u003c/em\u003e Micropropagation Ability of Black Chokeberry (\u003cem\u003eAronia melanocarpa\u003c/em\u003e (Michx.) Elliott). Horticulturae 9:1112. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/horticulturae9101112\u003c/span\u003e\u003cspan address=\"10.3390/horticulturae9101112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuran RE, Issah H (2022) The impact of strigolactone GR24 on \u003cem\u003eCapparis spinosa\u003c/em\u003e L. callus production and phenolic compound content. Plant Cell Tiss Organ Cult 149:197\u0026ndash;204\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEfferth T (2019) Biotechnology applications of plant callus cultures. Engineering 5:50\u0026ndash;59\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEkanayake IU, Meddage DPP, Rathnayake U (2022) A novel approach to explain the black-box nature of machine learning in compressive strength predictions of concrete using Shapley additive explanations (SHAP). Case Stud Constr Mate 16:e01059. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cscm.2022.e01059\u003c/span\u003e\u003cspan address=\"10.1016/j.cscm.2022.e01059\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEl Amri N, Chabir R, Kachkoul R, Bour A (2025) Phenolic Composition and Phytochemical Analysis of Moroccan Raw and Processed \u003cem\u003eCapparis spinosa\u003c/em\u003e Buds and Berries. Chem Biodivers. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cbdv.202500443\u003c/span\u003e\u003cspan address=\"10.1002/cbdv.202500443\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEren B, T\u0026uuml;rkoğlu A, Haliloğlu K et al (2023) Investigation of the influence of polyamines on mature embryo culture and DNA methylation of wheat (\u003cem\u003eTriticum aestivum\u003c/em\u003e L.) using the machine learning algorithm method. Plants 12:3261\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEsmaeilpour HR, Boostani R, Shoeibi A et al (2020) The potential effects of Caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) in the treatment of diabetic neuropathy. Trad Med Res 5:442\u0026ndash;448\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFahmideh L, Sheikhi M, Benakashani F, Solouki M (2019) Callus induction and organogenesis from various explants of plant \u003cem\u003eCapparis spinosa\u003c/em\u003e L. under \u003cem\u003eIn vitro\u003c/em\u003e conditions. J Plant Prod Res 26:75\u0026ndash;88. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.22069/jopp.2019.14326.2284\u003c/span\u003e\u003cspan address=\"10.22069/jopp.2019.14326.2284\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFallah Ziarani M, Tohidfar M, Navvabi M (2022) Modeling and optimizing \u003cem\u003ein vitro\u003c/em\u003e percentage and speed callus induction of carrot via Multilayer Perceptron-Single point discrete GA and radial basis function. BMC Biotechnol 22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12896-022-00764-4\u003c/span\u003e\u003cspan address=\"10.1186/s12896-022-00764-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFotiadou E, Cheilari A, Graikou K, Chinou I (2025) Comparative Study by HPTLC of Selected \u003cem\u003eCapparis spinosa\u003c/em\u003e Samples (Buds and Leaves) from the Cycladic Islands in Greece. Foods 14:1827\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGianguzzi V, Inglese P, Barone E, Sottile F (2019) \u003cem\u003eIn Vitro\u003c/em\u003e Regeneration of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. by Using a Temporary Immersion System. Plants 8:177. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants8060177\u003c/span\u003e\u003cspan address=\"10.3390/plants8060177\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoda SM, Ahmed SA, El Sherif F et al (2017) Genetically stable plants with boosted flavonoids content after \u003cem\u003ein vitro\u003c/em\u003e regeneration of the endangered \u003cem\u003eCapparis spinosa\u003c/em\u003e L. Glob. Drugs Ther 2:1\u0026ndash;7\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrimalt M, Hern\u0026aacute;ndez F, Legua P et al (2018) Physicochemical composition and antioxidant activity of three Spanish caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) fruit cultivars in three stages of development. Scientia hortic 240:509\u0026ndash;515\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrimalt M, Hern\u0026aacute;ndez F, Legua P et al (2022) Antioxidant activity and the physicochemical composition of young caper shoots (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) of different Spanish cultivars. Scientia Hortic 293:110646\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGristina AS, Fici S, Siragusa M et al (2014) Hybridization in \u003cem\u003eCapparis spinosa\u003c/em\u003e L.: Molecular and morphological evidence from a Mediterranean island complex. Flora 209:733\u0026ndash;741\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHagaggi NShA, Abdul-Raouf UM, Radwan TAA (2024) Variation of antibacterial and antioxidant secondary metabolites and volatiles in leaf and callus extracts of Phulai (\u003cem\u003eAcacia Modesta\u003c/em\u003e Wall). BMC Plant Biol 24:93. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12870-024-04747-9\u003c/span\u003e\u003cspan address=\"10.1186/s12870-024-04747-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHazrati S, Mousavi Z, Mollaei S et al (2025) Effect of Phenological Variation on the Phytochemical Composition and Antioxidant Activity of Different Organs of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. Horticulturae 11:702. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/horticulturae11060702\u003c/span\u003e\u003cspan address=\"10.3390/horticulturae11060702\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHesami M, Jones AMP (2021) Modeling and optimizing callus growth and development in \u003cem\u003eCannabis sativa\u003c/em\u003e using random forest and support vector machine in combination with a genetic algorithm. Appl Microbiol Biotechnol 105:5201\u0026ndash;5212. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00253-021-11375-y\u003c/span\u003e\u003cspan address=\"10.1007/s00253-021-11375-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHesami M, Jones AMP (2020) Application of artificial intelligence models and optimization algorithms in plant cell and tissue culture. Appl Microbiol Biotechnol 104:9449\u0026ndash;9485. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00253-020-10888-2\u003c/span\u003e\u003cspan address=\"10.1007/s00253-020-10888-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHigazy AE, El-Mahrouk ME, El-Banna AN et al (2023) Production of Black Cumin via Somatic Embryogenesis, Chemical Profile of Active Compounds in Callus Cultures and Somatic Embryos at Different Auxin Supplementations. Agronomy 13:2633. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/agronomy13102633\u003c/span\u003e\u003cspan address=\"10.3390/agronomy13102633\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIkeuchi M, Sugimoto K, Iwase A (2013) Plant callus: mechanisms of induction and repression. Plant Cell 25:3159\u0026ndash;3173\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeevan Ram PS, Singh S, Ali SA et al (2025) Machine Learning-Driven Screening of Promising Mutants in \u003cem\u003eNilgirianthus ciliatus\u003c/em\u003e Callus Cultures for Enhanced Phytochemicals Production. Physiol Plant 177:e70536. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/ppl.70536\u003c/span\u003e\u003cspan address=\"10.1111/ppl.70536\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarous O, Ben Haj Jilani I, Ghrabi-Gammar Z (2021) Ethnobotanical study on plant used by semi-nomad descendants\u0026rsquo; community in Ouled Dabbeb\u0026mdash;Southern. Tunisia Plants 10:642\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKatırcı R, Aasim M, Deveci G, Mustafa Z (2024) Comparing quantum machine learning and classical machine learning for \u003cem\u003ein vitro\u003c/em\u003e regeneration of cowpea (\u003cem\u003eVigna unguiculata\u003c/em\u003e). Plant Cell Tiss Organ Cult 159:32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-024-02880-9\u003c/span\u003e\u003cspan address=\"10.1007/s11240-024-02880-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaushik P, Rani M, Khurana N et al (2025) Revolutionizing Plant Tissue Culture: Harnessing Artificial Intelligence for Precision Propagation and Optimization. NPJ 15:e040624230642. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2174/0122103155302871240527094915\u003c/span\u003e\u003cspan address=\"10.2174/0122103155302871240527094915\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKianersi F, Abdollahi MR, Mirzaie-asl A et al (2020) Biosynthesis of rutin changes in \u003cem\u003eCapparis spinosa\u003c/em\u003e due to altered expression of its pathway genes under elicitors\u0026rsquo; supplementation. Plant Cell Tiss Organ Cult 141:619\u0026ndash;631\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirkan B, Ceylan O, SARIK\u0026Uuml;RKC\u0026Uuml; C, Bektas T (2021) Phenolic profile, antioxidant and enzyme inhibitory activity of the ethyl acetate, methanol and water extracts of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. Int J Secondary Metabolite 8:337\u0026ndash;351\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo\u0026ccedil;ak M, Yılmaz MC, Kuzğun C et al (2025) Synthetic seed production in \u003cem\u003eCrataegus monogyna\u003c/em\u003e L. and prediction of regeneration of synthetic seeds with machine learning algorithms. Plant Cell Tiss Organ Cult 161. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-025-03049-8\u003c/span\u003e\u003cspan address=\"10.1007/s11240-025-03049-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoufan M, Belkoura I, Mazri MA (2022) \u003cem\u003eIn Vitro\u003c/em\u003e Propagation of Caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.): A Review. Horticulturae 8:737. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/horticulturae8080737\u003c/span\u003e\u003cspan address=\"10.3390/horticulturae8080737\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLabbafi M, Mehrafarin A, Badi H et al (2018) Improve germination of caper (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) seeds by different induction treatments of seed dormancy breaking. Trakia J Sci 16:71\u0026ndash;74\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee KY, Nam D-H, Jeon Y et al (2024) Exploring the Production of Secondary Metabolites from a Halophyte \u003cem\u003eTetragonia tetragonoides\u003c/em\u003e through Callus Culture. Horticulturae 10:244. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/horticulturae10030244\u003c/span\u003e\u003cspan address=\"10.3390/horticulturae10030244\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu W, He Y, Xiang J et al (2011) The physiological response of suspension cell of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. to drought stress. J Med Plants Res 5:5899\u0026ndash;5906\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg SM, Erion G, Chen H et al (2020) From Local Explanations to Global Understanding with Explainable AI for Trees. Nat Mach Intell 2:56\u0026ndash;67. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s42256-019-0138-9\u003c/span\u003e\u003cspan address=\"10.1038/s42256-019-0138-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg SM, Lee S-I (2017) A Unified Approach to Interpreting Model Predictions\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa Q, Grones P, Robert S (2018) Auxin signaling: a big question to be addressed by small molecules. J Exp Bot 69:313\u0026ndash;328. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/jxb/erx375\u003c/span\u003e\u003cspan address=\"10.1093/jxb/erx375\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahood HE, Sarropoulou V, Tzatzani T-T et al (2024) Biomass and polyphenolic enhancement in date palm (\u003cem\u003ePhoenix dactylifera\u003c/em\u003e L.) callus cultures through \u003cem\u003eFusarium oxysporum\u003c/em\u003e elicitation. Plant Cell Tiss Organ Cult 159:76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-024-02934-y\u003c/span\u003e\u003cspan address=\"10.1007/s11240-024-02934-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoghadamnia Y, Kani SNM, Ghasemi-Kasman M et al (2019) The anti-cancer effects of capparis spinosa hydroalcoholic extract. Avicenna J med biotechnol 11:43\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohaddab M, El Goumi Y, Gallo M et al (2022) Biotechnology and \u003cem\u003eIn Vitro\u003c/em\u003e Culture as an Alternative System for Secondary Metabolite Production. Molecules 27:8093. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/molecules27228093\u003c/span\u003e\u003cspan address=\"10.3390/molecules27228093\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohaddab M, Genva M, Malika F et al (2024) \u003cem\u003eCapparis spinosa\u003c/em\u003e: A Rich Source of Phenolic Compounds-A Comprehensive Review of Its Phytochemistry, Health Benefits, and Biotechnological Applications. Biocatal Agric Biotechnol 103409\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMovafeghi A, HABIBI G, ALI AM (2008) Plant regeneration of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. using hypocotyl explants\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMunasinghe SP, Somaratne S, Weerakoon SR, Ranasinghe C (2020) Prediction of chemical composition for callus production in \u003cem\u003eGyrinops walla\u003c/em\u003e Gaetner through machine learning. Inf Process Agric 7:511\u0026ndash;522\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurashige T, Skoog F (1962) A Revised Medium for Rapid Growth and Bio Assays with Tobacco Tissue Cultures. Physiol Plant 15:473\u0026ndash;497. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1399-3054.1962.tb08052.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1399-3054.1962.tb08052.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoordijk B, Garcia Gomez ML, ten Tusscher KHWJ et al (2024) The rise of scientific machine learning: a perspective on combining mechanistic modelling with machine learning for systems biology. Front Syst Biol 4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fsysb.2024.1407994\u003c/span\u003e\u003cspan address=\"10.3389/fsysb.2024.1407994\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNowruzian A, Aalami A (2023) Effect of different treatments on dormancy breaking for seed germination enhancement and metabolite analysis of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. J Med plants By-Products 12:387\u0026ndash;396\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u0026Ouml;zcan E, Atar HH, Ali SA, Aasim M (2023) Artificial neural network and decision tree\u0026ndash;based models for prediction and validation of \u003cem\u003ein vitro\u003c/em\u003e organogenesis of two hydrophytes\u0026mdash;\u003cem\u003eHemianthus callitrichoides\u003c/em\u003e and \u003cem\u003eRiccia fluitans\u003c/em\u003e. \u003cem\u003eIn Vitro\u003c/em\u003e Cell Dev Biol - Plant 59:547\u0026ndash;562. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11627-023-10367-z\u003c/span\u003e\u003cspan address=\"10.1007/s11627-023-10367-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOzyigit II, Dogan I, Hocaoglu-Ozyigit A et al (2023) Production of secondary metabolites using tissue culture-based biotechnological applications. Front Plant Sci 14:1132555\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasternak TP, Steinmacher D (2024) Plant Growth Regulation in Cell and Tissue Culture. Vitro Plants 13:327. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants13020327\u003c/span\u003e\u003cspan address=\"10.3390/plants13020327\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeng S, Rajjou L (2024) Advancing plant biology through deep learning-powered natural language processing. Plant Cell Rep 43:208\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eP\u0026eacute;rez-Mej\u0026iacute;a N, Villarreal ML, S\u0026aacute;nchez-Carranza JN et al (2024) Phytochemical Profiles and Cytotoxic Activity of \u003cem\u003eBursera fagaroides\u003c/em\u003e (Kunth) Engl. Leaves and Its Callus Culture. Plants 13:1622. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants13121622\u003c/span\u003e\u003cspan address=\"10.3390/plants13121622\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetch J, Di S, Nelson W (2022) Opening the Black Box: The Promise and Limitations of Explainable Machine Learning in Cardiology. Cana J Cardiol 38:204\u0026ndash;213. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cjca.2021.09.004\u003c/span\u003e\u003cspan address=\"10.1016/j.cjca.2021.09.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahimi VB, Rajabian A, Rajabi H et al (2020) The effects of hydro-ethanolic extract of \u003cem\u003eCapparis spinosa\u003c/em\u003e (\u003cem\u003eC. spinosa\u003c/em\u003e) on lipopolysaccharide (LPS)-induced inflammation and cognitive impairment: Evidence from \u003cem\u003ein vivo\u003c/em\u003e and \u003cem\u003ein vitro\u003c/em\u003e studies. J Ethnopharmacol 256:112706\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRanade R, Joshi N, Kudale S (2023) Comparative secondary metabolite expression in callus cultures and mother plant in \u003cem\u003eBarleria prionitis\u003c/em\u003e L. Plant Cell Tiss Organ Cult 155:653\u0026ndash;663. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-023-02585-5\u003c/span\u003e\u003cspan address=\"10.1007/s11240-023-02585-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSadat-Hosseini M, Arab MM, Soltani M et al (2022) Predictive modeling of Persian walnut (\u003cem\u003eJuglans regia\u003c/em\u003e L.) \u003cem\u003ein vitro\u003c/em\u003e proliferation media using machine learning approaches: a comparative study of ANN, KNN and GEP models. Plant Methods 18:48. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13007-022-00871-5\u003c/span\u003e\u003cspan address=\"10.1186/s13007-022-00871-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSagharyan M, Ganjeali A, Cheniany M, Kouhi SMM (2020) Optimization of callus induction with enhancing production of phenolic compounds production and antioxidants activity in callus cultures of \u003cem\u003eNepeta binaloudensis\u003c/em\u003e Jamzad (\u003cem\u003eLamiaceae\u003c/em\u003e). Iran J Biotechnol 18:e2621\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarabandi M, Singh RK, Kalantari S et al (2024) Unveiling machine learning\u0026rsquo;s impact on \u003cem\u003ein vitro\u003c/em\u003e callogenesis optimization in \u003cem\u003eCrocus sativus\u003c/em\u003e L. S Afr J Bot 168:1\u0026ndash;8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarmah P, Borah T, Mudoi KD et al (2025) \u003cem\u003eIn vitro\u003c/em\u003e micropropagation of the critically endangered \u003cem\u003eAquilaria malaccensis\u003c/em\u003e Lam. and comparing the effect of PGRs\u0026rsquo; through machine learning models. Plant Cell Tiss Organ Cult 162:6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-025-03115-1\u003c/span\u003e\u003cspan address=\"10.1007/s11240-025-03115-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlimani C, El Goumi Y, Rais C et al (2021) Optimization of callogenesis/caulogenesis induction protocol in saffron plant (\u003cem\u003eCrocus sativus\u003c/em\u003e L.) using response surface methodology. Biointerface Res Appl Chem 12:4731\u0026ndash;4746\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSobhy FS, Ahmed S, Abdelrahim E et al (2025) Chemo Preventive Potential of Rutin-Rich Extract Obtained from \u003cem\u003eCapparis spinosa\u003c/em\u003e Culture against HepG2 Cancer Cell Line Activity. Egypt J Chem 68:855\u0026ndash;866\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSozzi GO, Peter KV, Babu KN, Divakaran M (2012) Capers and caperberries. Handbook of herbs and spices. Elsevier, pp 193\u0026ndash;224\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSozzi GO, Vicente AR (2006) Capers and caperberries. Handbook of Herbs and Spices. Elsevier, pp 230\u0026ndash;256\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSvolacchia N, Sabatini S (2023) Cytokinins. Curr Biol 33:R10\u0026ndash;R13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cub.2022.11.022\u003c/span\u003e\u003cspan address=\"10.1016/j.cub.2022.11.022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTan WH, Tong CY, Chua MX, Derek CJC (2024) Modelling and optimizing secondary metabolites production in \u003cem\u003eSpirodela polyrhiza\u003c/em\u003e using machine learning. J Clean Prod 478:143986. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jclepro.2024.143986\u003c/span\u003e\u003cspan address=\"10.1016/j.jclepro.2024.143986\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeoh SC, Subramaniam S, Chew BL (2023) The effects of 2, 4-dichlorophenoxyacetic acid on the induction of callus from cotyledon and hypocotyl explants of butterfly pea (\u003cem\u003eClitoria ternatea\u003c/em\u003e). Malays Appl Biol 52:61\u0026ndash;72\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang YT, Gan L, Liu W et al (2007) Research on the callus inducing and the cell growth and metabolism characteristics of \u003cem\u003eCapparis spinosa\u003c/em\u003e L. College of Life Sci and Technolo, Huazhong Univ of Sci and Technolo. Wuhan China 7:1779\u0026ndash;1783\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWijerathna-Yapa A, Hiti-Bandaralage J, Pathirana R (2025) Harnessing metabolites from plant cell tissue and organ culture for sustainable biotechnology. Plant Cell Tiss Organ Cult 162:55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11240-025-03180-6\u003c/span\u003e\u003cspan address=\"10.1007/s11240-025-03180-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin Y, He Y, Liu W et al (2014) The durative use of suspension cells and callus for volatile oil by comparative with seeds and fruits in \u003cem\u003eCapparis spinosa\u003c/em\u003e L. PLoS ONE 9:e113668\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYousefi E, Abedi M, Aghajanzadeh TA, Moreno DA (2025) Caper bush (\u003cem\u003eCapparis spinosa\u003c/em\u003e L.) bioactive compounds and antioxidant capacity as affected by adaptation to harsh soils. Sci Rep 15:11893. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-025-97298-4\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-97298-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZarbakhsh S, Shahsavar AR, Soltani M (2024) Optimizing PGRs for \u003cem\u003ein vitro\u003c/em\u003e shoot proliferation of pomegranate with bayesian-tuned ensemble stacking regression and NSGA-II: a comparative evaluation of machine learning models. Plant Methods 20:1\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13007-024-01211-5\u003c/span\u003e\u003cspan address=\"10.1186/s13007-024-01211-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang H, Ma ZF (2018) Phytochemical and pharmacological properties of \u003cem\u003eCapparis spinosa\u003c/em\u003e as a medicinal plant. Nutrients 10:116\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"plant-cell-tissue-and-organ-culture-pctoc","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pcto","sideBox":"Learn more about [Plant Cell, Tissue and Organ Culture (PCTOC)](https://www.springer.com/journal/11240)","snPcode":"11240","submissionUrl":"https://submission.nature.com/new-submission/11240/3","title":"Plant Cell, Tissue and Organ Culture (PCTOC)","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Capparis spinosa L., Plant tissue culture, Callogenesis, Machine Learning, Rutin","lastPublishedDoi":"10.21203/rs.3.rs-8619922/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8619922/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e \u003cem\u003eCapparis spinosa\u003c/em\u003e L. is a Mediterranean medicinal species of high economic value, yet its large-scale propagation and metabolite production remain constrained by conventional approaches. A full factorial design was used to evaluate the effects of four plant growth regulators, 6-benzylaminopurine, kinetin, 2,4-dichlorophenoxyacetic acid, and 1-naphthaleneacetic acid, on fresh weight gain from leaf explants. Data from twenty hormonal treatments were modeled using four machine learning algorithms: Random Forest, Gradient Boosting, Extreme Gradient Boosting, and second-degree polynomial regression. Random Forest provided the highest predictive accuracy. SHapley Additive exPlanations analysis identified 2,4-dichlorophenoxyacetic acid as the dominant factor driving callogenesis, with 6-benzylaminopurine exerting a secondary synergistic effect, whereas kinetin and 1-naphthaleneacetic acid showed minimal or inhibitory influence. Experimental validation confirmed the five best Random Forest\u0026ndash;predicted hormonal combinations, including the optimal mixture of 6-benzylaminopurine and 2,4-dichlorophenoxyacetic acid, which produced the highest increase in callus fresh weight gain. Rutin, the major bioactive flavonoid of \u003cem\u003eC. spinosa\u003c/em\u003e, was identified by LC-QTOF-MS/MS tandem mass spectrometry and semi-quantified by LC-TQ-MS/MS under 6-benzylaminopurine and 2,4-dichlorophenoxyacetic acid combinations. A stacked Random Forest model integrating fresh weight gain predictions successfully estimated rutin accumulation, with maximal production at moderate hormone levels. This integrative machine learning and SHapley Additive exPlanations framework offers an interpretable and scalable strategy for optimizing callus culture and enhancing high-value metabolite production in \u003cem\u003eC. spinosa\u003c/em\u003e. Moreover, callus culture represents a promising and sustainable alternative for large-scale production of valuable metabolites, reducing reliance on wild plant resources.\u003c/p\u003e","manuscriptTitle":"Integrating Machine Learning and SHAP Analysis to Boost Callus Growth and Rutin Biosynthesis in Capparis spinosa L.","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-23 12:03:17","doi":"10.21203/rs.3.rs-8619922/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2026-02-12T04:52:28+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-21T17:56:11+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-21T06:47:16+00:00","index":"","fulltext":""},{"type":"submitted","content":"Plant Cell, Tissue and Organ Culture (PCTOC)","date":"2026-01-19T08:12:29+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"plant-cell-tissue-and-organ-culture-pctoc","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pcto","sideBox":"Learn more about [Plant Cell, Tissue and Organ Culture (PCTOC)](https://www.springer.com/journal/11240)","snPcode":"11240","submissionUrl":"https://submission.nature.com/new-submission/11240/3","title":"Plant Cell, Tissue and Organ Culture (PCTOC)","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"c13c6574-ea4a-4fe9-ba2d-bb24f721123b","owner":[],"postedDate":"January 23rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-23T12:03:18+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-23 12:03:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8619922","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8619922","identity":"rs-8619922","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}