Optimization and analysis of microwave-assisted extraction of bioactive compounds from Mimosa pudica L. using RSM & ANFIS modeling

Abstract

The central composite rotatable design (CCRD) based response surface methodology (RSM) and adaptive neuro-fuzzy inference system (ANFIS) statistical methodology was used to design and identify highly efficient extraction process parameters to get high yield of bioactive compound from Mimosa pudica L. In general, many of the process parameters are need to be effectively involved to maximize the yield of bioactive compounds. In this relation, the independent process parameters such as methanol concentration (X ₁), microwave power (X ₂), irradiation temperature (X ₃) and irradiation time (X ₄) was chosen in the process of microwave assisted extraction (MAE). The observed studied parametes were produced effective results in the range of 60–85% methanol concentration, 15–25% microwave power, 40–60 °C irradiation temperature and irradiation time 10–15 min. Moreover the optimal yields of TPC and TFC are 635–640 mg gallic acid equivalents (GAE)/g and 61.53–61.76 mg rutin equivalents (RU)/g of extract, and their antioxidant activities are 68.7–72.6% DPPHsc, 76.1–76.86% ABTSsc and FRAB value of 65.24–66.94 µg mol (Fe (II))/g could be obtained for specific optimized process variables. Further, the bioactive compound mimopudine was identified through high performance liquid chromatography (HPLC) in the obtained extract.

Introduction

A vigorous search and screening of bioactive compounds from biological sources has received wide attention in the food and pharmaceutical industries due to their biological importance. It is well known that unstable molecules formed due to free radicals mainly caused by increased levels of reactive oxygen species (ROS) as a result of oxidative stress [1, 2]. The over load of such free radicals causes irreparable damage and may also be irreversible with time could lead to certain degenerative diseases or more specifically cellular diseases such as diabetes mellitus, coronary heart diseases, cancer, neurodegenerative diseases, etc., ROS is a by product of normal cellular metabolism and it is released by mitochondrial respiration [3, 4]. The excessive generation of ROS mainly caused by ultraviolet radiations, consumption of repeatedly heated vegetable oils, cigarette smoking, excessive consumption of alcohol, high doses of nonsteroidal anti-inflammatory drugs, ischemia–reperfusion injury, chronic infections, and inflammatory disorders [5]. Oxidative stress occurs when the balance between antioxidants and ROS are disrupted because of either depletion of antioxidants or accumulation of ROS [6]. Secondary metabolites of the plant possess antioxidant chemicals in the form of active compounds polyphenols, glycosides and steroids that inhibits the oxidation of other molecules by netutralizing the free radicals [7]. Among them, polyphenols are particularly attractive because of its biological properties including antioxidant activities and inhibition of reactive oxygen species generation [8]. An antioxidant effect can be from competitive consumption of the oxidant, thus sparing the target molecules being protected, and from quenching the chain reaction propagating free radical oxidation [9].

Mimosa pudica L. (Mimosaceae) is a creeping annual or perennial herb, it is referred to as touch me not, live and die, shame plant and humble plant is a prostrate or semi-erect subshrub of tropical America and Australia, also found in India. The extract of M. pudica L. used traditionally in the treatment of various ailments including headache, migraine, insomnia, diarrhea, dysentery, urogenital infections, fever, piles, and fistula [10, 11]. Further, M. pudica L. have already proven their therapeutic importance that is an antioxidant, antidiabetic, hyperglycemic, tumor, antidepressant, anticonvulsant, acetylcholinestrase inhibitory, anti-ophidian, anti-venom, wound healing and antifertility [12, 13]. Several studies revealed the phytochemical profile of M. pudica L. extract presence of polyphenols, alkaloids, non-protein amino acid (mimosine), flavonoids C-glycosides, sterols, triterpenes, tannins, and fatty acids [14]. The presence of these bioactive compounds is low so it is undoubtedly the most difficult technological operation in the separation and purification of these polyphenolic compounds in M. pudica L.

In recent times the extraction technologies for the cropping biologically active polyphenolic compounds from botanical materials with higher yield, lower extraction time and less solvent consumption is a major area of interest for many researchers and manufacturers. Several extraction techniques such as soxhlet, blending, sonication, ultra-sound assisted, microwave-assisted, pressurized liquid, and supercritical carbondioxide extraction has been employed to separate bioactive compounds from solid matrices in laboratories. In comparison to the classical techniques microwave assisted extraction has become one of the promising methods for extraction of bioactive compounds from plant sources [15]. It enables rapid extraction of solute from solid matrices, decreasing the solvent quantity. MAE technique is known for its short extraction time as increased solubility of analyte and its lower surface tension of extraction media, solvent viscosity for improved efficiency of extraction [16]. Plant secondary metabolites naturally occur in the cell wall or the cytoplasm. Microwave irradiation enables forced heating of solvent in the core of material that allows cell rupture. Cell disruption facilitates mass transfer of the solvent to the plant material and secondary metabolites from the plant to the solvent leading to effective extraction. Literature study reveals that microwave has been used to extract pectin from banana peel [17], orange peel [18], apple pomace [19], lime [20].

The focus of study is to extraction of polyphenolic compounds from M. pudica L. and analyze the effect of parameters. Several parameters such as solvent type, solvent concentration, particle size, microwave power, irradiation temperature and irradiation time etc., among these the chosen parameters methnol concentration, microwave power, irradiation temperature and irradiation time are having significant influence on the extraction yield of bioactive compounds. In addition, the selected methanol concentration is very important for extraction of bioactive compounds, the desired compound might be soluble in particular solvent and concentration. Temperature is one of important key factor for thermoliable compounds and antioxidant properties. Further, microwave power and irradiation time was very useful for penetrate/rupture plant cell in the sample as well as to reduce the time of experimental period. Under these perspectives, a statistical tool for analyzing optimum conditions seems to be very useful with minimization of time, utilization of solvent, and limiting the number of experiments by fixing the range of variables to achieve the highest yield of bioactive polyphenolic compounds.

Response surface methodology (RSM) is one of such combination of mathematical with statistical tool that can be used to measure the quantitative data and their interaction terms from analytical experiments to establish and simultaneously solve multi-variant equations [21, 22]. RSM is based on a polynomial equation to build a model between the dependent and these independent factors as well as to develop a symmetrical model that will predict and determine the experimental optimum conditions [8, 23]. Central composite rotatable design (CCRD) and Box-Behnken design (BBD) are commonly used designs for RSM to obtain an optimized extraction of bioactive compounds [24,25,26]. The CCRD is composed of a cube part which is a full factorial that determine main and interaction effects, and a star design (α) that quantifies main and quadratic effects, while the Box-Behnken design is composed of rotated lower-dimensional designs, and it estimates all linear effects, quadratic effects and 2-way interactions. It does not allow the reduction in design, being much less flexible than the CCRD [27]. They do not contain any corner points in the design space. The axial points are outside of the bounds of the design space box defined by the factorial part of the design which produces rotability. This allows the predicted response to be estimated with equal variance regardless of the direction from the center of the design space. The central composite rotatable design (CCRD) is a second order symmetrical model which has been applied by many researchers for the bioactive compounds and antioxidant extraction [28,29,30]. Further, ANFIS (Adaptive neuro-fuzzy inference system) model is employed to predict the obtained optimal responses using the computational power of neural network and high level human like thinking fuzzy systems that is capable of producing best results for nonlinear processes [31]. ANFIS is a new soft, intelligent computing technique used to study the interaction and nonlinear effect between the variables. In the present study, optimization of five levels (−2,−1, 0, +1, +2), four factors (solvent concentration, microwave power, irradiation temperature and irradiation time) of CCRD with RSM along with ANFIS model was adopted to obtain optimum conditions with respect of TPC, TFC and showing maximum antioxidant scavenging activities.

Materials and methods

Materials

The chemical 2,4,6-tripyridyl-s-triazine (TPTZ), Folin–Ciocalteu’s phenol reagent (FCR), rutin and Gallic acid were obtained from Himedia laboratories Pvt. Ltd. Mumbai, India. A sensitive colorimetric scavenger 2,2-diphenyl-1-picrylhydrazyl (DPPH) and 2,2′-Azinobis(3-ethylbenzothiazoline-6-sulphonic acid) radical cation (ABTS) assay adapted were from Sigma-Aldrich, MO, USA. Merck, Mumbai, India supplied Aluminium chloride, Sodium carbonate, Sodium hydroxide, Ferric chloride and analytical grade solvents were used in the study. M. pudica L. was collected from the foot hills of Western Ghats in and around Thaniparai Hills (longitude: 77.6497511° and latitude: 9.7256118°), Virudhunagar district, India during early winter season. The washing of the whole plant was done in a tap water repeatedly and the thoroughly washed plant was allowed to dry naturally for 72 h and was finely ground. Thus finely grounded powder filtered using 60 mesh screen were safely stored in a desiccators for further experiments. The fine powder was subjected to MAE using Microwave Extractor (CATA R) provided by Catalyst Systems (Pune, India) attached with a magnetron of 2450 MHZ, nominal power of 850 W and a reflex unit operating at 2455 MHz with 10 power levels, time controller, exhaust system, beam reflector and a stirring device. The system is operable at a maximum pressure of 30 bar depending on solvent composition, volume and working temperature. Analytical instruments, UV–Vis spectrophotometer (Varian, Cary 50), HPLC (Model LC-8, Shimadzu, Japan), Mixer grinder (Preethi Electronics, India) and Rotary vacuum dryer (Varian Rotavac) were used.

Selection of MAE solvent

The preliminary experiment’s main aim is to select a suitable solvent system that could identify the presence of highest content of TPC, TFC and possessing highest DPPH* radicals scavenging activity of M. pudica L. extract. Different solvent systems, six of them namely, chloroform, diethyl ether, methanol, acetone, ethyl acetate and n-hexane were used for this investigation. Each solvent system extraction process involved using 2 g of fine powdered M. pudica L. with 20 mL of constant solvent concentration (70% v/v in water) at fixed values of microwave power (20%), irradiation temperature (50 °C) and irradiation time (10 min).

Selection of relevant process variables

The four process variables i.e., solvent concentration (%), microwave power (%), irradiation temperature (°C) and irradiation time (min) were investigated in respect of dependent variables, such as, TPC, TFC, DPPHsc, ABTSsc and FRAP in the extract of M. pudica L. The aqueous methanol has been used as an extraction solvent on the basis of the preliminary experimental result as shown in Table 1. The selected four independent variables were investigated at five levels, coded as −2, −1, 0, +1 and +2 (as shown in Table 2) and CCRD design was applied to explore the optimal combination of extraction variables of M. pudica L. sample. The ranges of independent variables, such as 60–85% of methanol concentration, 15–25% microwave power, 40–60 °C of irradiation temperature and 10–15 min irradiation time were selected as per the preliminary experimental result for the highest content of TPC, TFC and DPPHsc.

Table 1 Preliminary selection of appropriate extraction solvent

Table 2 Experimental range of coded and actual values for central composite rotatable design (CCRD)

MAE of polyphenolic compounds and antioxidants

Based on the preliminary experimental investigation pre-selected optimal solvent was chosen to be used for further experiments in MAE of polyphenolic compounds and antioxidants from M. pudica L. The MAE was conducted using 2 g of accurately weighed fine powder of M. pudica L. along with 20 mL of pre-selected solvent and it was placed in an extraction vessel further it was kept into inside of microwave cavity. Sets of experiments followed by varying concentrations of methanol (60–85%), microwave power (15–25%), irradiation temperature (40–60 °C) and irradiation time (10–15 min) and operating parameters were set according to CCRD of RSM. MAE, extracts were filtered through Whatman No.1 filter paper and concentrated using rotary vacuum dryer at 40 °C to obtain TPC, TFC and antioxidant activities.

Measurement of total polyphenolic content (TPC)

Folin–Ciocalteu’s phenol reagent (FCR) according to the method described by Singleton et al. with slight modifications was adapted to determine TPC of the extract [32]. Briefly, a small volume of extract (0.3 mL) mixed with of Folin–Ciocalteu’s reagent (1.8 mL) were allowed to stand at room temperature for 5 min, followed by the addition of 1.2 mL sodium carbonate (7.5%, w/v) solution. The blank sample, 0.3 mL of distilled water replacing 0.3 mL of extract was also prepared. The mixture was allowed to stand for a further 60 min in the dark was examined in the spectrometer for absorbance at 765 nm. Gallic acid as a standard was subsequently used to express as mg gallic acid equivalents (GAE)/g sample.

Measurement of total flavonoid content (TFC)

The total flavnoid content of the extract was determined using the method described by Siddhuraju and Becker with slight modifications employing Aluminium chloride [33]. Briefly, a mixture consisting of small quantity (1 mL) of the extract, 0.3 mL of 5% (w/v) sodium nitrite solution and 4 mL of 80% (v/v) methanol were prepared in 5 min, and subsequently 0.3 mL of 10%(w/v) aluminium chloride solution was added and mixed. Subsequently an elapse of 6 min, 3 mL of 1 µL sodium hydroxide solution was added. After making up the volume of reaction mixture to 10 mL using distilled water, the resultant mixture was vortexed and absorbance was recorded at 510 nm. Using the standard curve prepared with rutin, TFC presence and its concentration was determined as mg rutin equivalents (RU)/g samples.

Determination of antioxidant power

%DPPH Scavenging assay

The method concerning the 2,2-diphenyl-1-picrylhydrazyl (DPPH) free radical scavenging activity reported by Brand-Williams et al. was used with slight modifications [34]. Aliquot of each sample extract (0.1 mL) was added to 3 mL of ethanolic solution of DPPH (0.1 µM). The mixture was shaken vigorously and kept in the dark for 30 min, and the absorbance was measured at 517 nm against a blank. Following the formula given below, capacity to scavenge the free radical DPPH in percentage of sample (%DPPH_SC) was calculated

$$\% {\text{DPP}}{{\text{H}}_{{\text{SC}}}}={\text{ }}\left( {{A_0} - {A_1}} \right){\text{ }} \times {\text{ }}100/{A_0}$$

(1)

where A ₀ = absorbance of the control; A ₁  = absorbance of the sample.

%ABTS scavenging assay

Following the method of Re et al. with some modifications ABTS* radical scavenging activity assay was carried out [35]. Incubation reaction between 7 mM ABTS (2,2′-Azinobis (3-ethylbenzothiazoline-6-sulphonic acid) diammonium salt) solution and 2.45 mM potassium persulphate solution in the dark maintained at room temperature was carried out for 16 h to generate ABTS*. Before spectrophotometry examination, ABTS solution was diluted with 0.3 mL ethanol and mixed with 0.1 mL of the extracts was thoroughly mixed vigorously and the absorbance at 734 nm was adjusted to 0.700 (±0.0020). Absorbance of the reaction mixture incubated for 6 min was examined at 734 nm using UV–Vis spectrophotometer. The %ABTS scavenged activity was calculated using the standard curve using rutin in 80% ethanol.

$$\% {\text{ ABT}}{{\text{S}}_{{\text{SC}}}}={\text{ }}\left( {{{\text{A}}_0} - {{\text{A}}_1}} \right){\text{ }} \times {\text{ }}100/{{\text{A}}_0}$$

(2)

where A ₀ = absorbance of the control; A ₁ = absorbance of the sample.

Ferric reducing antioxidant potential (FRAP) assay

Following Benzie and Strain and a method modified by Pulido et al. [36, 37]. FRAP antioxidant assay of the extract was carried out. Subsequently prepared FRAP reagent using 300 mM acetate buffer (3.1 g Sodium acetate, and 16 mL Acetic acid) at pH 3.6, 10 mM TPTZ (2,4,6-tripyridyl-s-triazine) solution in 40 mM hydrochloric acid solution, and 20 mM FeCl₃·6H₂O solution in distilled water was used to assess the activity. The acetate buffer (25 mL) and TPTZ (2.5 mL) were mixed together with FeCl₃·6H₂O (2.5 mL). Using the dark room, the plant extract (40 µL) maintained at 37 °C was allowed to react with the FRAP solution for 30 min and absorbance recorded at 593 nm. The standard curve was linear through 200 and 1000 µM FeSO₄. Results calculated in µM Fe (II)/g dry mass was compared with ascorbic acid as a standard.

Experimental design and optimization

The optimization of process variables using MAE of bioactive polyphenolic compounds was investigated using RSM based on CCRD. It is comprised of 30 experimental run with 16 and 8 factorial and axial points (α) respectively at a distance of ±2 from center and six replicates of central points are shown in Table 2.

The number of experiments was calculated from Eq. (3)

$${\text{N}}={\text{ }}{2^{\text{k}}}\,\left( {{\text{factorial points}}} \right){\text{ }}+{\text{ }}2{\text{k }}\left( {{\text{axial points}}} \right){\text{ }}+{\text{ }}{{\text{n}}_0}\,\left( {{\text{central points}}} \right)$$

(3)

where N is total number of experiments, k is the independent variable number, and n₀ is replicate number of the central points, which resulted in an experimental design of 30 points. Experimental data were fitted in the second order polynomial model to obtain the correlation between the dependent variable and independent variable. The model using response surface analysis is as follows:

$$Y={\beta _0}+\sum\limits_{{i=1}}^{3} {{\beta _i}} {X_i}+\sum\limits_{{i=1}}^{3} {{\beta _{ii}}} {X_i}^{2}+\sum\limits_{{i=1}}^{2} {\sum\limits_{{j=i+1}}^{3} {{\beta _{ij}}} } {X_i}{X_j}+\varepsilon$$

(4)

Based on the value of four variables, the Eq. (4) could converted as given below:

$$Y={\text{ }}{\beta _0}+{\beta _1}{X_1}+{\beta _2}{X_2}+{\beta _3}{X_3}+{\beta _4}{X_4}+{\beta _{12}}{X_1}{X_2}+{\beta _{13}}{X_1}{X_3}+{\beta _{14}}{X_1}{X_4}+{\beta _{23}}{X_2}{X_3}+{\beta _{24}}{X_2}{X_4}+{\text{ }}{\beta _{34}}{X_3}{X_4}+{\beta _{11}}X_{1}^{2}+{\beta _{22}}X_{2}^{2}+{\text{ }}{\beta _{33}}X_{3}^{2}+{\beta _{44}}X_{4}^{2}$$

(5)

where Y is the dependent variables (TPC (y ₁), TFC (y ₂), DPPH^*(y ₃), ABTS^* (y ₄) and FRAP(y ₅), β ₀ is the model constant, β _i, β _ii , and β _ij are model coefficients, X _i and X _j are coded value of independent variables, and ε is an error. Subsequently additional experiments were carried out to statistically verify the error in the process variables.

Optimization using adaptive neuro-fuzzy inference system (ANFIS) modeling

A hybrid system called “An adaptive neuro-fuzzy inference system or adaptive network-based fuzzy inference system (ANFIS) where the fuzzy system with expert knowledge stand as a front end pre-processor for the neural network input and output layers”. The knowledge based fuzzy system calculate the parameters using the neural network algorithms by utilizing the historical data thus generated. The solution for function approximation problems in a neural network platform is obtained through the data driven process. Fuzzy rules extracted from the input–output data set form the initial fuzzy model in the Fuzzy inference system. As a next step initial fuzzy model is built by fine tuning the rules adapted in the neural network and thus ANFIS methodology of the network is trained. Through this way an optimal data selection criterion shall be applied to reduce the number of training of data set. Thus a novelty of the technique adapted in the selection of input–output data pairs is advantageous.

In this study, same RSM experimental data used for ANFIS software to analyze the individual predicted output of TPC, TFC and antioxidant activities. The architecture of ANFIS used in this study was four input (methanol concentration, microwave power, irradiation temperature and irradiation time), and one output (TPC/TFC/DPPH/ABTS/FRAP) at a time (Fig. 1). The architecture of ANFIS topology architecture of feed-forward three-layered back propagation neural network is depicted in Fig. 1 and the ANFIS rule for effective extraction of active biomolecules, first determines the number of rules and antecedent membership functions and then uses linear least square estimation to determine each rule’s consequent equations. Consider a Sugeno Fuzzy Inference System (FIS), two inputs namely ‘x’ & ‘y’ and one output as ‘z’. A first order Sugeno FIS has the rules as following:

Fig. 1

The architecture of the ANFIS input and output model

• Rule 1: If x is A₁ and y is B₁, then f₁ = p₁x + q₁y + r₁

• Rule 2: If x is A₂ and y is B₂, then f₂ = p₂x + q₂y + r₂

The number of membership function assigned to each input variable is chosen by trial and error. Fuzzy logic tool box in MATLAB v R2013a was used for training, testing and validation of the network model for predicting the response for extraction of bioactive compounds and major antioxidants from M. pudica L.

HPLC analysis

Using BDS Hypersil RP-C 18 (Thermo, 5 µm, 120 Å, 250 mm × 4.6 mm) column maintained at temperature 25 °C and Rheodyne injector the optimally obtained extract was filtered and injected (20 µL) through a membrane filter (Millipore,USA). The mobile phase composed of methanol and water (60:40 vol%; pH 3.4) eluted at a flow rate of 1.0 mL/min and the effluent was monitored at 254 nm by UV detector. The obtained peak was detected and compared with the standard.

Statistical data analysis

The data collected from all the experiments were fed through the design expert (version 8.0.7.1, Stat-Ease, Inc., 2021 East Hennepin Ave, Suite 480, Minneapolis, MN 55413, USA). Three dimensional (3D) response surfaces and two dimensional (2D) contour plots were obtained through the application of optimal extraction conditions.

Result and discussion

Fitting the model

The extraction of bioactive polyphenolic compounds from M. pudica L. was carried out through computer controlled microwave assisted extractor. The experimental and model predicted values of TPC, TFC and showing antioxidant activities, namely, %DPPH, %ABTS, FRAP in M. pudica L. extracts obtained from all the experiments are shown in Table 3. According to the experimental design, the optimized results were obtained at methanol 85%, microwave power 25%, irradiation temperature 60 °C and irradiation time 15 min, under this circumstance, the optimal yields of TPC and TFC are 635 mg gallic acid equivalents (GAE)/g and 61.53 mg rutin equivalents (RU)/g of extract, and their antioxidant activities are 72.6% DPPHsc, 76.1% ABTSsc and FRAB value of 66.94 µg mol (Fe (II))/g.

Table 3 Central composite rotatable design with experimental responses and predicted responses

The coefficients of second order polynominal is shown in the Table 4 by fitting the experimental results in the quadratic model utilizing the data. The significance of the coefficients were explained by performing analysis of variance ANOVA statistical tool. F-test performed showed the significance of each coefficient. It could ascertained that if F value becomes greater and p value become smaller then the corresponding variables is said to be more significant [38]. Further p > 0.05 indicates the coefficient becomes statistically significant. The obtained F value (F = 39.34) and p value (p = 0.0004) is implied that the model was significant. The fitness and adequacy of the model was judged by the determination of multiple regression coefficients (R²) and the significance of lack-of-fit. The experimental results of the CCRD showed that the coefficients of determination (R²) of the models were 95% for TPC, TFC and antioxidant activities, suggesting that 95% of the actual levels can be matched with the model-predicted levels.

Table 4 Analysis of variance (ANOVA) for the quadratic polynomial mode

Analysis of the model

TPC

In Table 4 and Eq. (6) showed significant (p < < 0.05) contribution of the maximum extraction yield of TPC are the linear term of methanol concentration (X ₁), irradiation time (X ₄), and quadratic term X ₁ ², X ₂ ², X ₄ ². The correlation coefficient (R²) of microwave assisted extraction of TPC value in predicting model was 0.9202 with p value of lack of fit was 0.0004. The observed values are considered to be significant to establish model is a fitting one. The second–order polynomial equation for the fitted quadratic model for TPC in coded variables are given in below Eq. (6)

$${y_1}=577.5+101.12{X_1}+2.45{X_2}+11.45{X_3}+42.29{X_4}+1.43{X_1}{X_2}+14.06{X_1}{X_3} - 21.81{X_1}{X_4}+3.31{X_2}{X_3}+{\text{ }}10.18{X_2}{X_4}+6.56{X_3}{X_4} - 40.73X_{1}^{2} - 28.23X_{2}^{2} - 14.61X_{3}^{2} - 22.98X_{4}^{2}$$

(6)

Figure 2a shows that the normal percentage probability plot for studentized residuals of X _1, X ₂, X ₃ and X ₄ and these variants are normally distributed and have no deviation. Further, the predicted data against experimental data exhibited a higher R² value (0.9202) compared to RSM’s adj R² value (0.8458). Figure 2b shows the high values of regression coefficient (R² ≫ 0.8) considered as a good fit. Figure 2c, d show 3D response surface and 2D contour plot reveals significant effect of methanol concentration and irradiation time in maximizing yield of TPC with microwave power and irradiation temperature were held at a fixed level (zero level = 20%, 50 °C respectively). The yield of TPC value in M. pudica L. extracts of various experiments were represented in Table 3 varied from 202 to 635 mg gallic acid equivalents (GAE)/g. Lowest content of TPC yield was obtained at methanol 60%, microwave power 25%, irradiation temperature 60 °C and irradiation time 10 min while highest content was obtainable at methanol 85%, microwave power 25%, irradiation temperature 60 °C and irradiation time 15 min. Methanol concentration and irradiation time were playing an important role for the highest yield of TPC, increasing methanol concentration and decreasing irradiation time correlated to a higher content of TPC. Mild heating might soften the plant tissue, weaken the cell wall integrity and enhance the phytochemical solubility, allowing for more compounds to distribute to the solvent. However, prolonged microwave treatment at a higher temperature may induce phytochemicals degradation.

Fig. 2

Normal percentage probability plot for the studentized residuals for highest yield of TPC (a), Relationship between experimental and predicted value for highest yield of TPC (b), Response surface and contour plot showing the combined effects of methanol concentration (X ₁) and irradiation time (X ₄) for highest yield of TPC when microwave power and irradiation temperature were held at fixed level (zero level = 25%, 60 °C, respectively) (c) and (d)

TFC

The second order polynomial Eq. (7) and ANOVA Table 4 shows the significant (p ≪ 0.05) contribution of liner term X _1, X ₄, interaction term X ₁ X ₄ and quadratic term X ₁ ², X ₂ ², X ₃ ², X ₄ ² for maximum content of TFC in the M. Pudica L. extract. Also the response surface analysis of TFC content in the extract was demonstrated high regression coefficient value R² = 0.9624 and p value for lack of fit was 0.0168, these high values of regression coefficient (R² ≫ 0.8) indicate a good fit of experimental model. The second-order polynomial equation for the fitted quadratic model for TFC in coded variables are given in Eq. (7)

$${y_2}={\text{ }}55.78+10.14{X_1}+0.21{X_2}+1.74{X_3}+4.83{X_4}+0.63{X_1}{X_2}+0.62{X_1}{X_3} - 2.88{X_1}{X_4}+0.78{X_2}{X_3}+0.52{X_2}{X_4}+0.73{X_3}{X_4} - 3.09X_{1}^{2} - 3.73X_{2}^{2} - 1.93X_{3}^{2} - 2.08X_{4}^{2}$$

(7)

Figure 3a shows that the normal percentage probability plot of studentized residuals of X _1, X ₂, X ₃ and X ₄ and these variants are normally distributed and have no deviation. Figure 3b shows the high values of regression coefficient (R² ≫ 0.8) considered as a good fit. In addition, 3D response surface and the contour plot shown in Fig. 3c, d illustrate the effects of methanol concentration and irradiation time on the maximum content of TFC in the extract, when microwave power and irradiation temperature were held at a fixed level (zero level = 20%, 50 °C, respectively). Table 3 varied from 22.23 to 61.53 mg rutin equivalents (RU)/g. Lowest content of TFC yield was obtained at methanol 60%, microwave power 25%, irradiation temperature 60 °C and irradiation time 10 min while highest content was obtainable at methanol 85%, microwave power 25%, irradiation temperature 60 °C and irradiation time 15 min.

Fig. 3

Normal percentage probability plot for the studentized residuals for highest concentration of TFC (a), Relationship between experimental and predicted value for highest concentration of TFC (b), Response surface and contour plot showing the combined effects of methanol concentration (X ₁) and irradiation time (X ₄) for highest yield of TFC when microwave power and irradiation temperature were held at fixed level (zero level = 25%, 60 °C, respectively) (c) and (d)

Antioxidant activities (%DPPHsc, %ABTSsc and FRAP)

From the RSM analysis values in Table 3 and model equations (8)–(10) shows the linear term of methanol concentration (X ₁), irradiation time (X ₄), the interaction term of X ₁ X ₂ and followed by the quadratic term X ₁ ², X ₂ ², X ₃ ² has significant effects of all three antioxidant activities. Additionally, Table 4 and the respective model equations also indicated the linear term X ₃, interaction term X ₁ X ₄, X ₃ X ₄ and quadratic term X ₄ ² significantly influence %DPPH activity. Similarly, quadratic term X ₄ ² and interaction term X ₂ X ₃ has significance (p < 0.05) effect on FRAP. The correlation coefficient (R²) value of the models in %DPPHsc, %ABTSsc and FRAP are 0.9377, 0.9705 and 0.9261 respectively; with p value of lack of fit were 0.0122, 0.0031 and 0.0007 respectively. The observed value was signified that the model is a considerably fitting one. The second-order polynomial equation for the fitted quadratic models for %DPPH, %ABTS, FRAP in coded variables are given in Eq. (8)–(10).

$${y_3}={\text{ }}57.29+7.74{X_1} - 0.32{X_2}+1.79{X_3}+4.23{X_4}+1.98{X_1}{X_2}+1.06{X_1}{X_3}+2.51{X_1}{X_4}+0.72{X_2}{X_3}+0.86{X_2}{X_4}+1.69{X_3}{X_4} - 3.13X_{1}^{2} - 3.76X_{2}^{2} - 1.96X_{3}^{2} - 2.91X_{4}^{2}$$

(8)

$${y_4}={\text{ }}50.97+12.47{X_1}+1.03{X_2}+0.71{X_3}+6.62{X_4}+3.32{X_1}{X_2}+1.37{X_1}{X_3} - 1.27{X_1}{X_4}+0.64{X_2}{X_3}+0.44{X_2}{X_4}+0.99{X_3}{X_4} - 1.38X_{1}^{2} - 1.50X_{2}^{2} - 1.17X_{3}^{2} - 0.15X_{4}^{2}$$

(9)

$${y_5}={\text{ }}55.17+9.75{X_1}+0.47{X_2}+1.66{X_3}+5.08{X_4}+2.38{X_1}{X_2}+1.59{X_1}{X_3} - 0.35{X_1}{X_4}+3.02{X_2}{X_3} - 0.38{X_2}{X_4}+0.92{X_3}{X_4} - 3.50X_{1}^{2} - 3.36X_{2}^{2} - 3.32X_{3}^{2} - 3.63X_{4}^{2}$$

(10)

Figures 4a, 5a and 6a shows that the normal percentage probability plot of studentized residuals of X _1, X ₂, X ₃ and X ₄ and these variants are normally distributed and have no deviation. Further, Figs. 4b, 5b and 6b and the predicted data against experimental data for all three antioxidant activities gave a higher R² values (0.9377, 0.9705 and 0.9261 respectively) compared to RSM’s adj R² value (0.8795, 0.9431 and 0.8572 respectively) and the high values of regression coefficient (R² ≫ 0.8) is a good fit. The 3D response surfaces and 2D contour plot for antioxidant activities (%DPPHsc, %ABTSsc and FRAP) as a responsible functional variable of methanol concentration and irradiation time are shown in Figs. 4c, d, 5c, d and 6c, d. The figures show that methanol concentration of (85%), and irradiation time of (15 min) correspond to the highest antioxidant activities (%DPPHsc, %ABTSsc and FRAP). The maximum yields of antioxidant or antioxidant activities are %DPPH 72.6%, %ABTS: 76.1% and FRAP: 66.94 µg mol Fe (II)/g.

Fig. 4

Normal percentage probability plot for the studentized residuals for highest % DPPHsc activity (a), Relationship between experimental and predicted value for highest % DPPHsc activity (b), Response surface and contour plot showing the combined effects of methanol concentration (X ₁) and irradiation time (X ₃) for highest % DPPHsc activity microwave power and irradiation temperature were held at fixed level (zero level = 25%, 60 °C, respectively) (c) and (d)

Fig. 5

Normal percentage probability plot for the studentized residuals for highest %ABTSsc activity (a), Relationship between experimental and predicted value for highest % ABTSsc activity (b), Response surface and contour plot showing the combined effects of methanol concentration (X ₁) and irradiation time (X ₃) for highest % ABTSsc activity microwave power and irradiation temperature were held at fixed level (zero level = 25%, 60 °C, respectively) (c) and (d)

Fig. 6

Normal percentage probability plot for the studentized residuals for highest FRAP activity (a), Relationship between experimental and predicted value for highest FRAP activity (b), Response surface and contour plot showing the combined effects of methanol concentration (X ₁) and irradiation time (X ₃) for highest FRAP activity microwave power and irradiation temperature were held at fixed level (zero level = 25%, 60 °C, respectively) (c) and (d)

ANFIS modeling analysis

Adaptive neuro-fuzzy inference system modeling was performed to predict the extraction parameters of bioactive compounds present in the M. pudica L. using 30 experimental data presented in Table 3. Fuzzy logic toolbox in Matlab v R2013a was used to train the ANFIS and obtain the results. The examination was performed to select the set of optimal combination of parameter inputs with the maximum influence on the predictive accuracy of the output parameters. Accordingly the membership functions build in a fuzzy inference system (FIS) of ANFIS model consist of four inputs, five output (one at a time) achieved. The each input variable was defined as low, medium and high are the three fuzzy sets defined for each input variable such as solvent concentration, microwave power, irradiation temperature and irradiation time. Correspondingly, experimental data on predicted output responses were TPC (641 mg gallic acid equivalents (GAE)/g), TFC (61.47 mg rutin equivalents (RU)/g), DPPHsc (68.4%), ABTSsc (74.93%), and FRAP (65.45 µg mol (Fe (II))/g) were defined in five fuzzy sets namely very low, low, medium, high and very high. A total of 81 network nodes and 27 fuzzy rules were used to build the fuzzy inference system. The fuzzy rules were constructed on the basis of experimental data and human experiences. The predicted values of responses through RSM were also used to refine the fuzzy rules.

Validation of the model

The optimal extraction parameters were verified for the highest bioactive polyphenolic content (TPC, TFC) and showing maximum antioxidant activities (DPPH, ABTS, and FRAP) based on the values obtained using RSM. The experiment was carried out through Design Expert software, and it was able to identify optimum extraction parameters and their combinations. Further, the optimized parameters were also verified through ANFIS model using the same data. The experimental results showed that the methanol concentration and irradiation time had significant effects on the yields of bioactive polyphenolic compounds from M. pudica L. Applying optimum conditions based on combination of responses with minor variations experiments conducted is shown in Table 5. Based on those optimal conditions, methanol concentration of 85–87.5%, a microwave power 25%, irradiation temperature 60 °C and irradiation time 15 min. Under this condition while the experimental values of TPC, TFC, %DPPHsc, %ABTSsc and FRAP were 635–640 mg gallic acid equivalents (GAE)/g), 61.53–61.76 mg rutin equivalents (RU)/g, 68.7–72.6%, 76.1–76.8% and 65.24–66.94 µg mol (Fe (II))/g respectively, predicted values from RSM’s are TPC, TFC, %DPPHsc, %ABTSsc and FRAP were 642–643.04 mg gallic acid equivalents (GAE)/g), 62.29–62.64 mg rutin equivalents (RU)/g, 67.81–69.08%, 73.11–75.68% and 65.52–66.65 µg mol (Fe (II))/g respectively, Rule viewer plot (Fig. 7) provides the value of responses by varying the process variables. At the targeted optimized process conditions of process variables (methanol concentration = 87.5%, microwave power = 25% irradiation temperature = 60 °C, irradiation time = 15 min), the responses obtained through ANFIS model were TPC, TFC, %DPPHsc, %ABTSsc and FRAP were 641 mg gallic acid equivalents (GAE)/g), 61.47 mg rutin equivalents (RU)/g, 68.7, 74.93%, and 65.3 µg mol (Fe (II))/g respectively, in M. pudica L. extract. There exists a close fit between the obtained experimental values, the regression model and the predicted values obtained from RSM and ANFIS modelling.

Table 5 Verification of experimental and predicted values under optimum conditions based on combination of responses

Fig. 7

ANFIS rule viewer for the effect of process variables on responses for extraction of TPC, TFC and antioxidants from M.pudica L. extract

HPLC analysis

HPLC analysis of optimally obtained methanolic extract of M. pudica L. presented a distinct peak at a retention time of 9.141 which is similar to standard mimopudine (retention time 9.439) as shown in Fig. 8a and b. Few more peaks with varying retention times were also observed in this fraction.

Fig. 8

HPLC chromatogram of purified M. pudica L. extract (a), and mimopudine standard (b)

Conclusions

Optimization of extraction parameters in MAE of bioactive polyphenolic compounds and antioxidant activities of M. pudica L. extract was successfully achieved using RSM based on CCRD along with ANFIS modeling. The observed results showed that the independent parameters methanol concentration and irradiation time has significantly influenced the extraction yield of bioactive polyphenolic compounds. The interactions among the methanol concentration, microwave power, irradiation temperature and irradiation time, and quadratic terms were also significant effects on the yield of bioactive polyphenolic compounds. The optimized parameters were verified and proved by using fitting values of observed experimental values and predicted values. The second-order polynomial equations predicted by extraction conditions of the highest yield of TPC, TFC, %DPPHsc, %ABTSsc, and FRAP activities with methanol concentration of 85–87.5% v/v, microwave power 25%, irradiation temperature of 60 °C and irradiation time of 15 min. Mimopudine was the major bioactive polyphenolic compound present in the obtained M. pudica L. extract. RSM along with ANFIS optimization methods can be helpful for designing and extraction of other active compounds from the M. pudica L and other plant sources on industrial scale.