Antifungal activity and hair growth stimulation of purple sweet potato leaf fraction ( Ipomoea batatas (L.) Lamk) and its molecular mechanism through androgen receptor inhibition

activity and


Introduction
Hair loss, known medically as alopecia, is a dermatological condition affecting the scalp and often significantly impacting an individual's quality of life due to psychosocial consequences.The causes of hair loss are varied and encompass factors such as microbial (fungal) infections, genetic influences (hormonal), exposure to chemicals, medication usage, and environmental factors.Studies suggest that around 60 to 70% of the global population encounters androgenetic alopecia (AGA), primarily linked to excessive production of the androgen 5α-dihydrotestosterone (5α-DHT) in hair follicles, particularly in dermal papilla cells, which play a crucial role in regulating hair growth (Otberg et al. 2007;Azzouni et al. 2012).The binding of 5α-DHT to androgen receptors (AR) results in the down-regulation of androgen-sensitive genes in dermal papilla cells, leading to hair loss.Another approach to mitigating the impact of androgens in alopecia involves inhibiting the interaction between androgens and their receptors (Hamada et al. 1996;Inui and Itami 2011).
Officially recognized alopecia treatments include topical minoxidil and oral finasteride (Wu et al. 2016).Additionally, topical ketoconazole is also used for fungal-related hair loss resistant to minoxidil and finasteride (Okokon et al. 2015;Fields et al. 2020).However, prolonged use of these drugs may lead to potential long-term side effects and relapses upon discontinuation (Rossi et al. 2012;Wu et al. 2016).Hence, there's a growing interest in exploring natural alternatives for alopecia treatment.
The abundance of compounds in natural sources presents diverse pharmacological potentials, offering a promising avenue for drug discovery and development, especially in the field of hair care cosmetics for alopecia treatment (Mathur and Hoskins 2017;Atanasov et al. 2021).Among these natural sources, sweet potatoes (Ipomoea batatas (L.) Lamk), an herbaceous plant in the Convolvulaceae family, stand out for their potential hair care properties.Commercially, sweet potato root tubers are utilized, while the leaves serve as both vegetables and medicines (Islam 2014;Slamet and Andarias 2018).In Cameroon, sweet potato leaves are used for hair care by grinding, boiling, and macerating them before application (Fongnzossie et al. 2017).
Moreover, previous research has indicated that purple sweet potato leaves contain a variety of secondary metabolites, including steroids, terpenoids, alkaloids, polyphenols, tannins, flavonoids, vitamins, and minerals (Islam 2014;Sun et al. 2014;Ogunmoye et al. 2015;Lee et al. 2016;Dinu et al. 2018).The specific composition of these secondary metabolites may vary based on the region of cultivation and the extraction solvent employed.I. batatas leaves have demonstrated pharmacological activities, including antimicrobial, antidiabetic, antioxidant, anti-inflammatory, wound-healing, and properties associated with dengue haemorrhagic fever (Heyne 2007;Sun et al. 2014;Dinu et al. 2018).
In this study, our focus is on exploring the secondary metabolites, antifungal effects, and hair growth-stimulating potential of fractions from I. batatas leaves.Additionally, a computational study was conducted to comprehend the molecular mechanisms of metabolites from these plants toward the androgen receptor.This investigation is intended to provide scientific insights as an alternative for the development of natural-based treatments to stimulate hair growth affected by alopecia.

Collecting material and fractionation
The leaves of Ipomoea batatas were collected from Southeast Sulawesi, Indonesia.The I. batatas plant was identified by the School of Life Sciences and Technology, Bandung Institute of Technology, West Java, Indonesia, under registration number 6726/I1.CO2.2/PL/2019.The leaves underwent a series of preparations, including sample harvesting, wet sorting, water washing, deformation by cutting into small pieces, and subsequent drying in an Air Performance Ovens Frailabo® at 50 °C to obtain dry simplicial.The simplicia powder (7000 g) was macerated with ethanol as a solvent, and the solvent was changed daily over three days.It was then concentrated using a rotary evaporator (Buchi R-100) at 50 °C into a crude extract (711.34 g) with a yield of 10.16% (w/w).A total of 260 g of the ethanolic extract underwent fractionation through successive steps: it was dissolved in hot distilled water and then fractionated with n-hexane, ethyl acetate, and butanol solvents.This process yielded an n-hexane fraction of 50 g (19.23% (w/w)), an ethyl acetate fraction of 71 g (27.31% (w/w)), a butanol fraction of 74 g (28.46% (w/w)), and aqueous fractions of 50 g (19.23% (w/w)).The ethyl acetate fraction was selected for further analysis using LC-MS/MS.
Additionally, the identification of flavonoid compounds from the ethyl acetate fraction of I. batatas utilized the AC-QUITY UPLC BEH Shield RP18 column (100 × 2.1 mm, 1.7 µm) with eluent A (water + 0.1% formic acid) and eluent B (acetonitrile + 0.1% formic acid).The flow rate was set at 0.3 mL/minute, column temperature at 40 °C, injection volume of 3 µL, and both UV and MS detectors were employed.Post-processing of the samples was carried out using the MarkHerb database (EBM Scientific and Technology, Ltd).

Antifungal activity
The Malassezia furfur samples were obtained from Indonesia University, Jakarta.These fungi were maintained on a potato dextrose agar-olive oil (PDA-oil) medium.For three days, the fungal culture was grown in PDA oil at 35 °C.The fungal culture suspension was adjusted to a visually comparable turbidity of 0.5 MacFarland standard scales, which equated to 1.5 × 10 9 CFU/ml.The antifungal activity of the extract and its fractions was assessed using the agar diffusion method (Kaneko et al. 2005).Briefly, the PDA oil medium was prepared and placed in sterile Petri plates.The fungal suspension was uniformly spread on the surface of the PDA oil.A 20 μL suspension of the extract and its fractions was added to a sterile well at various concentrations.Sodium CMC suspension served as a negative control, while ketoconazole suspension was used as a positive control.The plates were then incubated at 35 °C for three days to observe the fungistatic effect.Three replicate plates were used for each treatment.

Hair growth activity
Two-month-old local rabbits, weighing approximately 2 kg, were utilized in this study.The rabbits underwent acclimatization to laboratory conditions and received standard feed and water.The photoperiod was set at 12 hours of light and 12 hours of darkness at room temperature (25 ± 3 °C).Following protocol approval from the Halu Oleo University Ethics Committee, all treatments on the tested animals were performed to minimize the number of animals and their suffering.Three male rabbits were selected for the study, and the Tanaka method with modifications was employed (Tanaka et al. 1980).
To prepare for the application of the extract and fraction, the dorsal area hairs on the rabbits' backs were shaved and treated with depilatory cream 24 hours before application.Six boxes, each measuring 2 × 2 cm 2 , were created with a spacing of 2 cm between them.Each rabbit received different doses of extract suspensions (5%, 10%, 20%, and 40%, respectively), while a vehicle control (1% Sodium CMC suspension) served as the negative control, and 2% minoxidil suspension acted as the positive control.The samples were applied to the test area box with 1 ml twice daily for 21 days.Hair growth stimulating activity was assessed by observing hair growth on the rabbit's back skin.
Hair length measurements began on the third day to evaluate the growth state of newly grown hair.Following the initiation of treatment, all rabbit hairs were selected for measurement, and the average length was calculated.The results were presented as the average hair length + SD (standard deviation) of 10 hairs.Tests were conducted to identify hair growth activity in the extract and its fractions.

Statistical analysis
The analyzed data were expressed as mean ± SD.The statistical analysis was performed using IBM SPSS Statistical software version 25.All data underwent one-way analysis of variance (ANOVA), and the significance level between treatment means was determined using the LSD range test at p < 0.05.

Molecular docking simulation
We obtained the three-dimensional (3D) structure of the androgen receptor (AR) from the Protein Data Bank (PDB ID: 4K7A) (https://www.rcsb.org/)(Hsu et al. 2014).The 3D structure of the I. batatas compound content, successfully identified via LC-MS/MS, was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/).AutoDockTool 1.5.6 was utilized to prepare both the receptor and ligands.After removing water molecules, the receptor underwent processing by adding Kollman charge and protonations.Concurrently, the ligand structure received Gasteiger charge and was configured to rotate freely (Morris et al. 2009;Arfan et al. 2022).The docking process was executed using AutoDock Vina software (Oleg and Arthur J. 2010).The binding site was set based on the minoxidil position as the native ligand on the x, y, and z-axis, encompassing a grid area of 27 × 27 × 27 Å.Other docking parameters were adjusted following default steps.Subsequently, the interactions between each ligand and AR were analyzed using Discovery Studio Visualizer software.

Molecular dynamics simulation
The four best compounds from the docking results were then proceeded to the molecular dynamics simulation stage using GROMACS 2016 (Abraham et al. 2015).The AR was parameterized using the AMBER99SB-ILDN force field (Petrov and Zagrovic 2014).Ligand parametric files were generated using ACPYPE (Sousa Da Silva and Vranken 2012).To neutralize the system, Na and Cl ions were introduced, and solvation was modeled employing the TIP3P water model.Pressure and temperature were set at 1 atm and 310 K, respectively.The Particle Mesh Ewald technique was employed for long-range electrostatic force calculations (Wang et al. 2016).System stability was confirmed by assessing root-mean-square deviation and fluctuations (RMSD and RMSF), and radius of gyration (Rg).Binding affinities of each ligand with AR were also calculated using the MM/PBSA method, facilitated by the g_mmpbsa package (Kumari et al. 2014).

Design expert
The Design-Expert 10 (DX-10) software was utilized to optimize the proportions of excipients in the hair tonic formulation of the ethyl acetate fraction from I. batatas leaves.A D-optimal mixture design was employed with the constraint ethanol 96% (X 1 ) + propylene glycol (X 2 ) + Tween 80 (X 3 ) = 30%, as depicted in Table 1.The component ranges were set as follows: 5 ≤ X 1 ≤ 15, 10 ≤ X 2 ≤ 24, and 1 ≤ X 3 ≤ 5. Design-Expert software generated 16 runs, with 11 runs being different and five runs as replicates (7 runs had no replicates, 3 runs had 2 replicates, and 1 run had 3 replicates).Through the D-optimal mixture approach, the effects of these components on the hair tonic's physicochemical properties were assessed, and the optimal combination was determined.For optimization, the combination of factors that yielded the best responses was determined based on the effect of each factor.

Hair tonic preparation
The formulation of the hair tonic is outlined in Table 1.Initially, the ethyl acetate fraction is combined with ethanol and stirred until homogeneous (solution 1).Subsequently, propylene glycol is added to solution 1 (solution 2).Tween 80 is then introduced into solution 2 (solution 3).DMDM hydantoin and sodium metabisulfite are dissolved in deionized water to create solution 4. Following this, solutions 3 and 4 are mixed.Finally, deionized water is added to reach the volume limit, and the mixture is stirred until it achieves homogeneity.

Measurement of responses (physicochemical properties)
The pH was measured using a pH meter calibrated by dipping the electrode into two solutions, assuming that the pH of the test solution fell between the pH values of the two solutions.Commonly used solutions for calibration are pH 4 and pH 7 (Patil et al. 2018).Viscosity was determined using the Ostwald viscometer, measuring the time required for the liquid to pass through two marks as it flowed through a vertical capillary tube (Apriani et al. 2021).Density was determined using the pycnometer method, measuring the weight of distilled water and hair tonic in the pycnometer (Febriani et al. 2016).

Statistical and data analysis
Fitting response values were conducted using special quartic and quadratic models (Eqs.1-2).ANOVA at P < 0.05 was used to assess the statistical significance of the equation.
Y represents the predictive dependent variable (responses), such as physicochemical properties like pH, viscosity, and density.λ denotes constant coefficients for model terms, and X represents the proportions of real-components.

Phytochemical analysis
Ipomoea batatas has traditionally been used for hair care, specifically addressing issues like hair loss or alopecia.The causes of alopecia vary from genetic to environmental factors.In light of this traditional application, a phytochemical screening was conducted to identify its phytochemical compounds and assess its antifungal and hair growth stimulant activities in this study.The ethyl acetate subfractions from I. batatas were analyzed using the LC-MS/MS method (Suppl.material 1).This analytical technique was chosen for its known selectivity, sensitivity, and accuracy in rapid analysis.LC-MS/MS analysis revealed that the subfractions of I. batatas contained twenty-three compounds.Additionally, the LC-MS/MS analysis identified thirteen probable compounds such as steroids (stigmastane-3-6-dione), alkaloids (pyropheophorbide A and methyl-pyropheophorbide A), flavonoids (hyperoside, quercetin, and kaempferol), terpenoids (digiprolactone, α-onocerin, and tussilagonone), amide (moupinamide), phenolics (3-hydroxy-4-methoxy-cinnamic acid), and fatty acids (triacontanoic acid and trichosanic acid) (Table 2).

Antifungal activity
The in-vitro antifungal assay of the extract revealed varying degrees of activity against Malassezia furfur, depending on the concentration.Specifically, at a 5% ethanolic extract concentration, no discernible antifungal activity was observed.However, at a concentration of 10%, the extract exhibited a mild level of antifungal activity, which further intensified with increasing concentrations (Fig. 1A and Table 3).Similarly, the assessment of the fraction unveiled distinct antifungal properties across different fractions against M. furfur.Notably, strong antifungal activity was observed in the aqueous and ethyl acetate fractions, while the n-hexane and butanol fractions demonstrated weaker antifungal effects (Fig. 1B and Table 4).
While providing valuable initial insights, this study is considered a preliminary test.Further investigations are imperative to delve deeper into the antifungal potential of the extract and fractions.It is essential to identify the secondary metabolite compounds responsible for the observed antifungal effects.Alkaloids, terpenoids, flavonoids, tannins, and polyphenols are among the secondary metabolites believed to contribute to antifungal activity and merit closer scrutiny in subsequent studies.The findings from this research lay the groundwork for future exploration, offering a promising avenue for deriving antifungal activity from I. batatas leaves.
The mechanism of alkaloids as antifungals is to disrupt the formation of the fungal cell membrane.Alkaloids bind to ergosterol, creating holes that lead to cell membrane leakage.This leakage results in fungal cell damage and eventual cell death (Dhamgaye et al. 2014).Similarly, tannins act by binding to the cell membrane structure of the fungus.Tannins have an affinity for ergosterol and polyphenols, and their binding to fungal membranes may contribute to antifungal action (Carvalho et al. 2018).
Polyphenols play a crucial role in plants, contributing to resistance against microorganisms, herbivores, and insects (Hammerschmidt 2005).Phenolics or polyphenols, characterized by a benzene ring with one or more hydroxyl groups, exhibit increased toxicity when a hydroxyl group is present on the benzene ring (Ruiz-García and Gómez-Plaza 2013).Various mechanisms of phenolics have been suggested to counteract pathogenic microbes, including disrupting enzymatic processes, affecting energy production and synthesis of structural components, weakening and destroying the cell membrane's permeability barrier, altering cells' physiological status, or affecting nucleic acid synthesis (Cutter 2000).
Flavonoid compounds may act through several targets, such as forming complexes with proteins through nonspecific bonds like hydrogen bonds and hydrophobic effects, and by forming covalent bonds.The mechanism of antimicrobial action may be associated with their ability to attach to microbes, cell envelope transport proteins, enzymes, and other targets.Lipophilic flavonoids may also interfere with microbial membranes (Cowan 1999;Mishra et al. 2009).Meanwhile, terpenoids exhibit antifungal activity through interference with cell wall permeability and inhibiting the formation of pseudohyphae and chlamydoconidia (Leite et al. 2015).

Hair growth activity
The concentration of the ethanol extract in stimulating hair growth yielded different results at all concentrations compared to 2% minoxidil.Similar outcomes were also observed with the I. batatas fraction (Figs 2, 3).The research results indicate a significant increase in hair growth after the administration of 2% minoxidil, the ethanol extract, and the fraction of I. batatas leaves (Tables 5, 6).However, at 5% and 10% ethanol extract concentrations, the butanol and water fractions exhibited a less effective A B i j k a g h hair growth profile compared to 2% minoxidil.Meanwhile, ethanol extract concentrations of 20% and 40%, as well as the n-hexane and ethyl acetate fractions, demonstrated effectiveness equivalent to 2% minoxidil, as they showed no significant differences based on statistical test results (Tables 5, 6).Based on these findings, it is evident that the ethanol extract of I. batatas leaves at concentrations of 20% and 40%, as well as the n-hexane fraction and ethyl acetate fraction, present greater opportunities as an alternative in treating hair loss.
The potential stimulation of hair growth is attributed to the secondary metabolite content of this plant.Phytochemical compounds, whether individually or in a mixture form, are often reported as non-medical treatments to alleviate symptoms of various dermatological hair conditions (Daniels et al. 2019).Alkaloids, such as caffeine, are known to play a role in modulating hair growth (Daniels et al. 2019).The presence of polyphenols offers opportunities for hair growth, as they are potent antioxidant and anti-inflammatory agents.Additionally, terpenoids and tannins also have the potential to act as hair growth stimulants through their action as 5α-reductase inhibitors (Kaushik et al. 2011;Jahan et al. 2020).Meanwhile, flavonoid compounds stimulate nerves and the production of IGF-1 (Harada et al. 2007).

Molecular docking analysis
Molecular docking analyses were utilized to predict the preferred binding modes of compounds from I. batatas leaves identified through LC-MS/MS results with the androgen receptor (AR).This involved binding affinity predictions and the identification of residual amino acid interactions.The objective of this analysis was to estimate the plant's potential to stimulate hair growth by inhibiting androgen receptors.It aimed to predict the binding energy, indicating the plant's inhibitory effect on androgen receptors and its ability to promote hair growth.Lower binding  2.1 ± 0.1 3.5 ± 0.5 5.2 ± 0.5 6.5 ± 0.2 7.7 ± 0.5 9.4 ± 0.5 13.83 + 0.6Â queous fraction 2.3 ± 0.2 4.5 ± 0.3 5.5 ± 0.5 6.9 ± 0.3 8.0 ± 0.5 9.5 ± 0.3 14.1 + 0.5N ote: * The mean difference is significant with negative control at the 0.05 level; ^ The mean difference is significant with 2% minoxidil at the 0.05 level.energy suggests higher binding efficiency and enhanced inhibition (Citra et al. 2023).Our focus included key amino acid residues like Tyr857, Gln858, Lys861, Glu793, Trp796, and Leu797 in the androgen receptor, crucial for ligand interaction (Hsu et al. 2014).The molecular docking process follows the procedure outlined by Henny et al. on the same receptor.This procedure yields an RMSD value of 1.64 Å, classified as valid, aiming to reproduce the minoxidil conformation on AR (Kasmawati et al. 2022a(Kasmawati et al. , 2022b)).Table 5 illustrates that the docking scores of 10 bioactive compounds in I. batatas leaves were lower than those of minoxidil and other compounds, ranging from -4.9 to -7.6 kcal/mol.Interestingly, the groups of flavonoid compounds (hyperoside and quercetin) as well as alkaloids (pyropheophorbide A and methyl-pyropheophorbide A) exhibit highly promising binding affinities to AR compared to minoxidil.These four compounds, have binding energies of -7.6, -7.5, -7.2, and -6.5 kcal/mol for hyperoside (1), pyropheophorbide A (2), methyl-pyropheophorbide A (3), and quercetin (4), respectively.Meanwhile, compounds 11 and minoxidil showed identical affinity energy.In contrast, compounds 12 and 13 demonstrated a more positive binding energy compared to the other compounds, with values of -4.5 kcal/mol and -3.8 kcal/mol, respectively.
Based on the energy analysis, we focused on examining the patterns of the four best compounds (Fig. 3).Compound 1 forms four hydrogen bonds with Arg786, Glu793, Asp864, and Leu862 residues.Additionally, hydrophobic interactions with the AR binding pocket's Lys861 and Asp864 residues were identified.Compounds 2 and 3 exhibit two hydrogen bonds with residues Arg786 and Ser865, and Glu793 and Leu862, respectively.Furthermore, these two compounds engage in hydrophobic interactions with Trp796, His789, Lys861, Leu797, and Pro868.Notably, in compound 4, only one hydrogen bond and hydrophobic interaction with Ser865 and Arg786 res-idues were observed.This analysis reveals that the four best compounds can interact with key residues of the AR while showcasing their ability to bind to this receptor.

RMSD analysis
The RMSD of the AR backbone, when complexed with the top four compounds from I. batatas fraction, is depicted in Fig. 4A, in comparison to minoxidil.In a system simulation, RMSD gauges the folded protein structure against the partially or entirely unfolded structure.It signifies the dynamic changes in the protein throughout the simulation, crucial for assessing protein stability (Sargsyan et al. 2017).According to the simulation results, all the top compounds exhibited superior stability compared to minoxidil, as indicated by the movement of the AR backbone residue.The average RMSD

RMSF analysis
RMSF assesses the fluctuations in the central carbon atom of the protein structure, representing the coordinate oscillations of each amino acid around its reference point during dynamic simulation (Shao et al. 2022).The RMSF profiles of each of the top four compounds with AR reveal intriguing fluctuations (Fig. 5A).Intense oscillations were noted in the Gln670 and The918 residues in the N-terminus and C-terminal regions of the AR.Meanwhile, other residues exhibit comparable RMSF trends throughout the simulation.This analysis suggests that the top four compounds do not alter the flexibility of the amino acid residues in AR.

Gyration radius analysis
Protein compactness was assessed throughout the simulation by measuring the radius of gyration (Rg) for each complex.The Rg of a protein signifies the dispersion of its atoms around its axial direction, representing the distance between the rotating point and the point where energy transfer has the most significant impact.The interaction of a protein with a ligand can influence protein folding and stability, and this influence can be tracked through their patterns and Rg values.A stable protein folding behavior during the simulation is characterized by low and consistent Rg values (Sneha and George Priya Doss 2016).As depicted in Fig. 5B, the AR complexed with compounds 3 and 4 exhibited an impact on protein folding around the ~45 ns simulation time, which then stabilized again, similar to other compounds, until the end of the simulation.Notably, compound 1 demonstrated the lowest Rg value with an average of 1787 nm.In contrast, compounds 2-4 and minoxidil displayed similar Rg values, with averages of 1,795 nm, 1,795 nm, 1,796 nm, and 1,793 nm, respectively.

MM-PBSA Calculation
The molecular dynamics simulation facilitated the computation of binding energies for all identified compounds using the MM-PBSA method, as outlined in Table 8.MM/ PBSA methodology not only enables the quantification of total binding energy but also allows for the dissection of  individual energies contributing to specific details of the system binding.This approach sheds light on dominant interactions in the binding process, providing valuable insights for drug design strategies (Wang et al. 2019).The calculated total binding energy (∆E Bind ) for the compound 3 system was -66.862 kJ/mol, representing a lower value compared to the other compounds in the system.Compounds 2, 1, and 4 also exhibited robust binding affinity, with energies of -60,862 kJ/mol, -55,534 kJ/mol, and -39,695 kJ/mol, respectively.
Key contributors to the favorable binding across all systems were identified as van der Waals (∆E VDW ), electrostatic (∆E Ele ), and solvent-accessible surface area (∆E SASA ) energies.These factors played crucial roles in establishing favorable binding interactions.Conversely, polar solvation energy (∆E PB ) exhibited less favorable characteristics, particularly in compounds 1 and 2, where the values were positive compared to other compounds, indicating an unfavorable impact on the binding affinity of the androgen receptor (AR) complex.This observation aligns with the stability analysis of all the identified compounds, reinforcing their potential to inhibit the activity of the AR receptor.
Overall, the four best compounds from purple sweet potato leaves exhibit better binding energy than minoxidil based on energy calculations using the MM-PBSA method, which aligns with the docking results.The findings from this simulation study reinforce the potential of methyl pyropheophorbide A and hyperoside from this plant in binding with AR, contributing to their activity against hair loss.

Design expert result
The application of mixture design in pharmaceutical product development is an efficient method for optimizing formulation composition and gaining fundamental insights into the underlying relationship between independent and observed (dependent) variables (Basalious et al. 2010).In mixture design, the sum of all independent variables is kept constant, and only the relative proportions of each component are examined for their impact on the overall properties of the mixture (Martinello et al. 2006).While there are several types of mixture designs, such as simplex-centroid and simplex-lattice designs, D-optimal designs were chosen in this study for optimizing excipients when there are unusual restrictions on experimental settings.The criteria depend on minimizing the overall variance of the predicted regression coefficient by maximizing the value of the determinant of the information matrix (Karoui et al. 2023).
The best model was selected based on low standard deviation, low predicted sum of squares, and high R-squared p-values.The p-values of the acceptable models were lower than 0.05, and the p-values of lack of fit were higher than 0.05 (Amini Sarteshnizi et al. 2015).Following these response criteria, a special quartic model was the most suitable for pH results, while quadratic models were fitting for viscosity and density parameters.
Three independent variables (96% ethanol, propylene glycol, and tween 80 concentrations) were chosen, along with three response variables comprising hair tonic pH, viscosity, and density, to optimize physicochemical properties.Hair tonic is widely used to promote hair growth and strengthen hair follicles.Using excipients that can dissolve and enhance the penetration of active ingredients while maintaining good physicochemical properties is a crucial factor in hair tonic formulations.Ethanol, propylene glycol, and tween 80 play a role as solvents and penetration enhancers.
Sixteen experimental runs were conducted using Design Expert, and each product's pH, viscosity, and density were determined as responses.The results of the experiments are shown in Table 9, where the formulation variables include: X 1 : 96% ethanol; X 2 : propylene glycol; X 3 : tween 80, and the dependent variables include: Y 1 : pH, Y 2 : viscosity, Y 3 : density.
Tables of ANOVA can be applied to assess how well the model and each parameter fit the data by analyzing the mean least square error estimates to the mean pure experimental error and ensuring that the errors are nor-  mally distributed.Then, the F-test can be used to evaluate the significance of the fit for both the model and the individual parameters (Jeirani et al. 2012).The F-test can determine whether parameters are significant or non-significant.This check applies a 95% confidence interval to assess the significance of the parameter.Therefore, a probability value (P-value) of 5% would be a significant level in the F-test for interpreting results (Pertiwi et al. 2023).This study utilized ANOVA to identify which factors significantly affect pH, viscosity, and density.The statistical parameters applied in selecting and evaluating the best fitted model are regression data (p value and F value), lack-of-fit, coefficient of determination (R 2 ), adjusted coefficient of determination (adjusted R 2 ), and prediction coefficient of determination (prediction R 2 ).Statistical analysis also builds the most suitable model equation (Table 10) As shown from Table 10, for the pH model it is statistically significant (P_0.05);The overall model F-value of 8.53 and p-value of <0.05 illustrate that the model is considerable model terms.The p-value of the quartic model is 0.0098 (<0.05), which means that the results are significant.The lack of fit for pH is 0.5061 (>0.05).This value indicates good because there is no pH variation for each replication in the optimization process (Pertiwi et al. 2023).A negative predicted R 2 implies that overall means a better predictor of this response than the current model (Anderson et al. 2017).
Contour plots (Fig. 6) for each response were meticulously crafted utilizing the advanced features of Design-Expert software version 10 to achieve the utmost desirability in the optimization process.In-depth statistical scrutiny, detailed in Table 12, entailed a comprehensive one-sample test analysis to discern the optimal hair tonic based on prediction values generated by the software.The predictive value emanating from the solution exhibiting the highest desirability was carefully selected, attaining a specific balance with an exemplary desirability score of 0.881.This desirability was meticulously determined, considering the intricate interplay of ethanol, propylene glycol, and tween 80 (Fig. 6).The formulation of the hair tonic was intricately devised, aligning with the optimal levels dictated by the software's sophisticated algorithms.Upon practical implementation, the observed response remarkably mirrors the predictive value derived from the meticulously optimized process, as briefly presented in Table 13.Impressively, the observed results fall seamlessly within the anticipated prediction index range, affirming the robustness and reliability of the optimization strategy employed.

Conclusion
This research successfully demonstrated the potential of the ethanolic extract and its fractions from I. batatas leaves as an anti-alopecia agent by stimulating hair growth in a rabbit model and exhibiting antifungal activity by inhibiting the growth of M. furfur.The activity of the ethanolic extract and its fractions is attributed to secondary metabolite compounds.Alkaloid compounds (methyl pyropheophorbide A and pyropheophorbide A) and flavonoids (hyperoside and quercetin) identified in this plant have anti-alopecia activity based on in-silico studies through molecular simulation inhibiting androgen receptors.Based on their binding affinity predicted from MM-PBSA calculations and their stability during molecular dynamics, all four compounds showed better affinity and stability than minoxidil.Additionally, 96% ethanol, propylene glycol, and tween 80 at proportions of 13.39%, 13.71%, and 2.9%, respectively, could enhance the physicochemical properties of the hair tonic according to the D-optimal mixture design.The discoveries related to stimulating hair growth, antifungal properties, and the potential inhibition of the androgen receptor in this plant offer promising opportunities for its advancement as a therapeutic solution for hair loss associated with alopecia.

Figure 2 .
Figure 2. The hair growth profile of male rabbits (A) administered varying concentrations of ethanol extract and (B) fractions from I. batatas leaves.

Figure 4 .
Figure 4.The (A) RMSD backbone and (B) RMSD ligand analyses of the four best compounds from I. batatas against AR during molecular dynamics simulation.

Figure 5 .
Figure 5.The (A) RMSF and (B) radius of gyration analyses of the four best compounds from I. batatas against AR during molecular dynamics simulation.

Table 1 .
Mixture composition in hair tonic formulation with ethanol 96%, propylene glycol, and tween 80 in a three-component constrained D-optimal mixture design.

Table 2 .
The LC-MS/MS profile of the ethyl acetate subfraction from I. batatas leaves.
* The analysis of flavonoid content utilizing the MarkHerb database.

Table 3 .
Inhibition zone of ethanolic extract from I. batatas leaves against Malassezia furfur.

Table 4 .
Inhibition zone of fractions from I. batatas leaves against Malassezia furfur.

Table 5 .
Hair growth activity of ethanolic extracts from I. batatas leaves on male rabbits.

Table 6 .
Hair growth activity of ethanolic extract and fractions from I. batatas leaves on male rabbits.

Table 7 .
Binding affinity prediction of minoxidil and identified compounds from I. batatas leaves against AR.

Table 8 .
The binding energies of all system during 100 ns simulation.

Table 9 .
The results of each response from 16 runs of actual design by software DX-10.

Table 10 .
Regression coefficients and statistic data fitted from ANOVA for the adjusted model to experimental data in D-optimal mixtures design for physicochemical properties of hair tonic.significant at 0.05 level; p < 0.05 is significant; p > 0.05 is not significant. *

Table 11 .
Lower and upper limit of response.

Table 12 .
Desirability and formula solution issued from Design Expert software.

Table 13 .
Optimized values obtained by the confirmation.