Optimization of the Drum Drying Parameters and Citric Acid Level to Produce Purple Sweet Potato (Ipomoea batatas L.) Powder Using Response Surface Methodology

Purple sweet potato (PSP) is a rich source of anthocyanins, but the anthocyanin content and color can be affected by the drying method and processing condition. Response surface methodology (RSM) with a Box–Behnken design (BBD) was used to investigate the effects of citric acid (CA) concentration, steam pressure (SP) and rotation speed (DS) on the physicochemical and functional properties of drum-dried purple sweet potato powder (PSPP). The anthocyanins of the PSPP were analyzed using mass spectrometry with electrospray ionization and twelve anthocyanins were identified. The results indicated that the moisture content (4.80 ± 0.17–9.97 ± 0.03%) and water activity (0.290 ± 0.004–0.47 ± 0.001) (p < 0.05) decreased with increasing drum temperature as well as with reduced drum rotating speed. CA had a significant (p < 0.05) effect on the color and total anthocyanin content (101.83 ± 2.20–124.09 ± 2.89 mg/100 g) of the PSPP. High SP and low DS negatively affected the antioxidant properties of the PSPP. DPPH value of the PSPP ranged from 20.41 ± 0.79 to 30.79 ± 1.00 μmol TE/g. The optimal parameters were achieved at 0.59% CA, 499.8 kPa SP and 3 rpm DS.


Introduction
Sweet potato, known as Ipomoea batatas L., is one of the most important food crops globally [1] and is the sixth most important food crop after wheat, rice, maize, potatoes and cassava. Sweet potato is the fifth most important food crop in developing countries, playing a significant role in food security, with an annual global production of more than 105 million metric tons, of which 90 million tons are produced in the Asian region [2].
There are different varieties of sweet potato of various skin and flesh colors ranging from white, yellow, orange and light purple to deep purple. Purple-fleshed sweet potatoes are more popular among researchers due to the high concentration of anthocyanins that contributes to good antioxidant and other biological activities [2,3]. It has been reported that a sweet potato tuber has the sixth highest antioxidant content among 11 roots and tubers and it is ranked 14th among 22 vegetables [4,5]. Gras et al. [6] stated that the anthocyanin content of purple sweet potato (PSP) might vary depending on the variety, and they found that it ranged from 558 to 2477 mg/100 g DM for the studied freeze-dried high-anthocyanin PSP provenances. Furthermore, Nevara et al. [7] mentioned that the anthocyanin content of drum-dried PSP powder varied from 83.72 to 121.71 mg/100 g depending on the (98%) and delphinidin chloride (96%) were obtained from Chem Faces (Wuhan, China). Folin-Ciocalteu reagent, 2,2-diphenyl-1-picrylhydrazyl (DPPH) and gallic acid (98%) were obtained from Merck Company (Darmstadt, Germany). Trolox and 2,4,6-tripyridyl-striazine (TPTZ) were provided by Acros Organics (Geel, Belgium). Other chemicals and solvents used were of analytical grade bought from Fisher Scientific (Leicestershire, UK).

Sample Preparation
The unpeeled PSP tubers (~4 kg) were washed thoroughly, weighed and cut into cubes of approximately 3 cm (length) × 3 cm (width) × 1.5 cm (height) using a stainless-steel knife to facilitate a quick and uniform heat transfer during the steaming. Then, the PSP cubes were steamed at 100 • C for 30 min in a stainless-steel steam cooker, cooled to room temperature prior to pureeing using a bowl cutter (Mainca CM-21, Spain) for 10 min at low speed. Citric acid powder (0.5 to 1.5% wet basis) was added to the puree after one minute of mashing.

Drum Drying Operation
The acidified PSP puree was passed through a preheated double drum dryer (R. Simon Dryers Ltd., Nottingham, England). Preliminary studies and available literature were used to set steam pressure between 300 and 500 kPa and drum rotation speed from 1 to 3 rpm. Drum-dried PSP flakes were collected, immediately packed into laminated aluminum foil bags and sealed. Then, the dried flakes were milled using a high-speed stainless-steel grinder (IKA M20, IKA Labor-Technik, Staufan, Germany) for 20 s and sieved through a 425-µm mesh screen to obtain the PSP powder, which was sealed in laminated aluminum foil bags and stored at −20 • C for further analysis.

Experimental Design and Data Analysis
The experiment design was generated using RSM with the three-factor BBD to investigate the effects of three independent variables, namely, of citric acid (X 1 ), steam pressure (X 2 ) and drum rotation speed (X 3 ), on the physicochemical and functional properties of the PSPP. Table 1 displays the independent variables and respective coded and uncoded values. The polynomial regression equation used to investigate the dependent variable (Y) involves the main, interaction and squared effects on the response surface as shown in Equation (1): Y = β 0 + β 1 X 1 +β 2 X 2 + β 3 X 3 + β 4 X 1 X 2 + β 5 X 1 X 3 + β 6 X 2 X 3 +β 7 X 2 1 + β 8 X 2 2 + β 9 X 2 3 (1) where, Y is the response calculated using the models; β 0 , β 1 , β 2 , β 3 , β 4 , β 5 , β 6 , β 7 , β 8 and β 9 represent the regression coefficients; and X 1 , X 2 and X 3 are the independent variables for citric acid, steam pressure and drum speed, respectively. The moisture content of the PSPP was determined using the oven drying method [31] and expressed as a dry basis using the following formula: where, W 1 is the initial weight of the aluminum dish and the sample; W 2 is the final weight of the aluminum dish and the sample; and W 3 is the weight of the aluminum dish.

Determination of Water Activity
The PSPP's a w was determined using an AquaLab water activity meter (Model CX2, Pullman, DC, USA) [24].

Determination of Color
The color attributes of the PSPP were measured using a colorimeter (CR-410, Konica Minolta, Japan) based on the color coordinates L* (lightness), a* (redness) and b* (yellowness) at room temperature. A white tile was used to calibrate the instrument. The hue angle and chroma were calculated according to Nevara et al. [7] using the following formulas:

Determination of the Water Solubility Index and Water Absorption Capacity
The water solubility index (WSI) and water absorption capacity (WAC) of the PSPP were evaluated following the method described by Chang et al. [32] with minor modifications. One gram of the PSPP was mixed with 10 mL distilled water in a centrifuge tube in a water bath at 37 • C for 30 min before centrifugation at 3000 g for 10 min. The resulting supernatant was dried in a laboratory oven (Memmert, Büchenbach, Germany) at 105 • C for 24 h to obtain a constant dry solid weight and the weight of the precipitate at the bottom of the centrifuge tube was recorded. The WSI and WAC were calculated using the following equations: WSI = residual supernatant s dry weight dried powder s weight × 100 (5) WAC = weight of precipitate dried powder s weight -residual supernatant s dry weight × 100 (6) 2.5.5. Thermal Properties A differential scanning calorimeter (DSC) (Mettler DSC 823e, Mettler Toledo, Spain) was used to measure the thermal properties of the PSPP. About 3-5 mg powder was mixed with distilled water (1:3) and sealed in an aluminum pan. Then, the hermetically sealed pan was heated from 30 to 115 • C at a rate of 10 • C/min. An empty pan was used as a reference. Thermal properties including To (onset temperature), Tp (peak temperature), Tc (conclusion temperature) and ∆H (enthalpy change) were recorded [33]. The pH value of the PSPP was determined using a calibrated pH meter (Mettler Toledo, Switzerland). Briefly, 2 g of PSPP was mixed with 20 mL of distilled water and the electrode was placed in the solution. pH value was taken at room temperature (28 ± 2 • C).
2.6. Determination of the Antioxidant Activity 2.6.1. Extraction of Antioxidants The antioxidants were extracted from the PSPP as described by Yang et al. [34] with minor modifications. Briefly, 0.5 g of the PSPP sample were mixed with 15 mL of 80% methanol in a 50-mL centrifuge tube, vortexed for 30 s and placed in the dark for 2 h in a water bath fitted with a shaker at room temperature (28 ± 2 • C). Then, the tubes were centrifuged at 3000 g for 10 min and the supernatant was collected and stored in the dark at 4 • C in capped centrifuge tubes for further analysis.

UV Absorption Spectra Scanning
The collected extract (400 µL) was mixed with 3 mL potassium chloride buffer (pH 1.0) and the absorption spectrum was scanned from 200 nm to 700 nm using a double-beam UV-1650PC spectrophotometer (Shimadzu, Kyoto, Japan).
For HPLC quantification of PSPP anthocyanins, the anthocyanins were acid-hydrolyzed under high temperature. About 1 g PSPP was mixed with 10 mL 6 mol/L HCl and 30 mL 80% methanol in a tube with a screw cap. The tubes were sealed tightly and heated in a water bath at 90 • C for 90 min. The tubes were immediately cooled and brought to 50 mL with 80% methanol. The samples were centrifuged at 3000 g for 15 min and the supernatant was collected. Standards of cyanidin and peonidin at 0.2 to 0.0125 mg/mL were also hydrolyzed in the same manner [36]. HPLC analysis was performed using a Waters 2695 HPLC system equipped with a Waters 2487 dual λ absorbance detector on an RP-18 LiChrospher column, 5 µm particle size, 250 × 4.6 mm (Merck, Darmstadt, Germany). The solvents used were A, 5% (v/v) formic acid in water, and B, 100% methanol. Separation was achieved using gradient elution from 15 to 35% (B) in the first 15 min, keeping 60% B at 30 min, and reaching 80% (B) at 40 min and 15% at 45 min. The injection volume was 20 µL and the flow rate was 1.0 mL/min [37].

Determination of the Total Anthocyanin Content
The total anthocyanin content (TAC) was determined using the pH differential method as described by Jiang et al. [9] with some modifications. Briefly, 0.4 mL extract was mixed with 2.6 mL potassium chloride buffer (pH 1.0) and 2.6 mL sodium acetate buffer (pH 4.5) and incubated in the dark for 30 min at room temperature. The absorbance was measured at 527 nm and 700 nm using a UV-1650PC spectrophotometer (Shimadzu, Kyoto, Japan) and distilled water was used as a reference. The total monomeric anthocyanin content was expressed as cyanidin-3-glucoside equivalent using the following equation: where, A = (A 527 − A 700 ) pH 1.0 − (A 527 − A 700 ) pH 4.5 , MW = 449.2 g/mol (molecular weight of cyanidin-3-glucoside), V = volume of the extract (mL), L = 1 cm (cell path length), DF = dilution factor, ε = 26,900 (molar absorption coefficient of cyanindin-3-glucoside) and m = weight of the sample (g).

Determination of the Total Flavonoid Content
The total flavonoid content (TFC) was determined using the method of Yea et al. [38] with some modifications. One milliliter of the sample extract was mixed with 4.0 mL distilled water; then, 300 µL of 5% (w/v) NaNO 2 and 300 µL of 10% (w/v) AlCl 3 were added and the solution was allowed to stand for 5 min. After this, 2.0 mL of 1 M NaOH were added. The catechin solution (0-300 mg/L) was used to plot a standard curve and the absorbance was measured at 510 nm using a UV/Vis spectrophotometer (Shimadzu, Kyoto, Japan). The TFC of the PSPP was expressed in milligrams of catechin equivalent per gram (mg CE/g) of the sample.

Determination of the Total Phenolic Content
The total phenolic content (TPC) was determined using the Folin-Ciocalteu assay with some modifications [39]. Firstly, 0.5 mL of the extract or 0-300 µg/mL of the gallic acid solution were mixed with 0.5 mL of the Folin-Ciocalteau reagent and vortexed for 10 s. After 3 min, 2.0 mL of the 7.5% (w/v) sodium carbonate solution and 2.0 mL distilled water were added and the mixture was vortexed for 10 s and left in the dark for 2 h. The absorbance was determined at 760 nm using a UV-1650PC UV/Vis spectrophotometer. The results were expressed in milligrams of gallic acid equivalent per gram (mg GAE/g) of the sample.

Determination of DPPH Radical Scavenging Activity
The DPPH scavenging activity of the PSPP was assayed according to the procedure described by Brand-Williams et al. [40] with some modifications. Firstly, 0.1 mL of the extracted solution were mixed with 3.9 mL of the 80-µM DPPH solution, vortexed for 10 s and left in the dark for 30 min. The absorbance was measured at 517 nm using a UV-1650PC spectrophotometer and the radical scavenging activity was derived from the standard curve of the Trolox solution (0-1000 µM). The results were reported as µmol of Trolox equivalent per gram of the sample (µmol TE/g).

Determination of Ferric Reducing Antioxidant Power
The ferric reducing antioxidant power (FRAP) was analyzed according to Benzie et al. [41] with some modifications. A freshly prepared FRAP reagent (2.7 mL) was mixed with 30 µL PSPP extract or Trolox solution (0-2000 µM). Then, the mixture was vortexed for 10 s and incubated in the dark for 30 min before reading the absorbance at 593 nm using a UV-1650PC spectrophotometer. The results were reported as µmol of Trolox equivalent per gram of the sample (µmol TE/g).

Statistical Analysis and Validation of the Model
The samples were analyzed in triplicate and the data obtained were analyzed using RSM with the BBD to fit the second-order polynomial equation generated using Minitab version 17.0 (Minitab Inc., State College, PA, USA). Multiple regressions were used to correlate the independent variables to the response variables and the regression coefficients of the final models were determined. The significance and quality of the fit of the model were analyzed using analysis of variance (ANOVA). The optimization of drum drying con- ditions was performed using the numerical multiple optimization procedure to determine the most desirable physicochemical and functional properties of the drum-dried PSPP.
A validation test was conducted to determine the adequacy of the final reduced model and provide a recommendation for the optimized variables [42]. Then, the predicted optimum drying condition was compared with the experimental value of the response. Finally, the predicted and experimental values of each response were compared by one sample t-test to determine the validity of the model. If there was no significant difference (p > 0.05) between the predicted values and the experimental data, the final reduced model was considered valid.

UV Absorption Spectra of the PSPP
The UV spectra of PSPP extracts from 200-700 nm are shown in Figure 1. Woodall et al. [43] stated that, generally, anthocyanins exhibit two absorption peaks, ranging from 270 to 290 nm and 500 to 550 nm. However, acylation of anthocyanins with aromatic organic acids gives another absorption peak in the range of 310-320 nm. In this study, four typical peaks were observed in the UV spectrum, the first peak was observed at 290-297 nm, followed by 302-306 nm and 324 and 332 nm. The final peak was observed at approximately 527 nm in agreement with the results reported by Jing Li et al. [10]. According to Jie Li et al. [12] the PSP anthocyanins are more stable at low pH value ranging from 3 to 4, under this condition the acylated anthocyanins are more stable to heat treatment, pH changes and exposure to light. Figure 1 showed that the UV spectra of the extract of drum-dried PSP added with 1.5% CA have a higher absorbance compared to the control sample, indicating that CA can be used to retain anthocyanins during drum drying of PSP. by one sample t-test to determine the validity of the model. If there was no significant difference (p > 0.05) between the predicted values and the experimental data, the final reduced model was considered valid.

UV Absorption Spectra of the PSPP
The UV spectra of PSPP extracts from 200-700 nm are shown in Figure 1. Woodall et al. [43] stated that, generally, anthocyanins exhibit two absorption peaks, ranging from 270 to 290 nm and 500 to 550 nm. However, acylation of anthocyanins with aromatic organic acids gives another absorption peak in the range of 310-320 nm. In this study, four typical peaks were observed in the UV spectrum, the first peak was observed at 290-297 nm, followed by 302-306 nm and 324 and 332 nm. The final peak was observed at approximately 527 nm in agreement with the results reported by Jing Li et al. [10]. According to Jie Li et al. [12] the PSP anthocyanins are more stable at low pH value ranging from 3 to 4, under this condition the acylated anthocyanins are more stable to heat treatment, pH changes and exposure to light. Figure 1 showed that the UV spectra of the extract of drumdried PSP added with 1.5% CA have a higher absorbance compared to the control sample, indicating that CA can be used to retain anthocyanins during drum drying of PSP.

Response Surface Analysis
The effects of three independent variables on moisture content (MC), water activity (aw), L*, a* and b* values, hue (Hu) and chroma (Ch), WSI, water absorption capacity (WAC), pH, TAC, TPC, TFC, radical scavenging activity (DPPH) and FRAP of the drumdried PSPP are presented in Table 2. The independent and dependent variables were fitted to second-order polynomial as in Equation (1) to check for the goodness of fit. The result showed that the models were accurately fitted at the 95.0% confidence level, hence, can be used to predict the responses as a function of three independent variables studied (Tables 2 and 3). All the reduced models in Table 4

Response Surface Analysis
The effects of three independent variables on moisture content (MC), water activity (aw), L*, a* and b* values, hue (Hu) and chroma (Ch), WSI, water absorption capacity (WAC), pH, TAC, TPC, TFC, radical scavenging activity (DPPH) and FRAP of the drumdried PSPP are presented in Table 2. The independent and dependent variables were fitted to second-order polynomial as in Equation (1) to check for the goodness of fit. The result showed that the models were accurately fitted at the 95.0% confidence level, hence, can be used to predict the responses as a function of three independent variables studied (Tables 2 and 3). All the reduced models in Table 4 had significant (p < 0.05) p-values for regression with a high R 2 values (80.2-100%), showing the adequacy of the model fitting.   Regression coefficient (Coef); x 1 , x 2 , x 3 are dependent variables; moisture content (MC), water solubility index (WSI), water absorption capacity (WAC), total anthocyanin content (TAC), total phenolic content (TPC), total flavonoid content (TFC), radical scavenging activity (DPPH) and ferric reducing antioxidant power (FRAP) of the purple sweet potato powder (PSPP).

Model Adequacy
The final fitted reduced models were checked for the adequacy of approximation with a real system. The adequacy of the models was judged by residuals from the leastsquares fit and normal probability plots were used to check the normality assumption by constructing the residuals against normal present probability (Figure 2a). Based on the normal probability plots for all responses in this experiment, the normality assumption was fulfilled as the residual stand along a straight line in the normal probability plots [44]. In the versus fit plots, the scattered residuals on both sides of the line proved that the variances of the observations were constant for all the independent variables (Figure 2b). Based on this finding, both the normal probability plot and versus fits plot for all the responses were satisfactorily; hence, the experiential models were adequate to describe the RSM applied in this experiment.
In the versus fit plots, the scattered residuals on both sides of the line proved that the variances of the observations were constant for all the independent variables (Figure 2b). Based on this finding, both the normal probability plot and versus fits plot for all the responses were satisfactorily; hence, the experiential models were adequate to describe the RSM applied in this experiment.  Table 2 shows the main effect of three independent variables on MC and aw and the MC of the PSPP ranged from 4.80 to 9.97%. Based on Table 4, the effect of CA and DS was significantly positive (p < 0.05) on MC, whereas that of SP was negative (p < 0.05). The DS and SP had significant positive and negative effects, respectively, on aw, ranging from 0.290 to 0.473. The quadratic effect of CA was negatively related to both MC and aw, whereas all the other quadratic and interaction effects had an insignificant effect (p > 0.05) on MC and aw. As shown in Figure 3, the water activity and moisture content of the PSPP decreased rapidly with the increase in steam pressure, followed by a decrement in drum speed.  Pua et al. [21] also reported the comparable effect of drum drying parameters on aw and MC of the drum-dried jackfruit powder. In this study, reduction of water activity and moisture content of the PSPP with the increase in steam pressure at 300, 400 and 500 kPa was due to an increase of the drum surface temperature to 104.67 ± 2.05, 114.67 ± 2.05 and 128.00 ± 1.63 °C, respectively. Similarly, Kakade et al. [26] highlighted that increasing steam pressure amplifies the drum surface temperature and increases moisture evapora-  Table 2 shows the main effect of three independent variables on MC and a w and the MC of the PSPP ranged from 4.80 to 9.97%. Based on Table 4, the effect of CA and DS was significantly positive (p < 0.05) on MC, whereas that of SP was negative (p < 0.05). The DS and SP had significant positive and negative effects, respectively, on a w , ranging from 0.290 to 0.473. The quadratic effect of CA was negatively related to both MC and a w, whereas all the other quadratic and interaction effects had an insignificant effect (p > 0.05) on MC and a w . As shown in Figure 3, the water activity and moisture content of the PSPP decreased rapidly with the increase in steam pressure, followed by a decrement in drum speed.

Moisture Content and Water Activity
In the versus fit plots, the scattered residuals on both sides of the line proved that the variances of the observations were constant for all the independent variables (Figure 2b). Based on this finding, both the normal probability plot and versus fits plot for all the responses were satisfactorily; hence, the experiential models were adequate to describe the RSM applied in this experiment.  Table 2 shows the main effect of three independent variables on MC and aw and the MC of the PSPP ranged from 4.80 to 9.97%. Based on Table 4, the effect of CA and DS was significantly positive (p < 0.05) on MC, whereas that of SP was negative (p < 0.05). The DS and SP had significant positive and negative effects, respectively, on aw, ranging from 0.290 to 0.473. The quadratic effect of CA was negatively related to both MC and aw, whereas all the other quadratic and interaction effects had an insignificant effect (p > 0.05) on MC and aw. As shown in Figure 3, the water activity and moisture content of the PSPP decreased rapidly with the increase in steam pressure, followed by a decrement in drum speed.  Pua et al. [21] also reported the comparable effect of drum drying parameters on aw and MC of the drum-dried jackfruit powder. In this study, reduction of water activity and moisture content of the PSPP with the increase in steam pressure at 300, 400 and 500 kPa was due to an increase of the drum surface temperature to 104.67 ± 2.05, 114.67 ± 2.05 and 128.00 ± 1.63 °C, respectively. Similarly, Kakade et al. [26] highlighted that increasing steam pressure amplifies the drum surface temperature and increases moisture evapora- Pua et al. [21] also reported the comparable effect of drum drying parameters on a w and MC of the drum-dried jackfruit powder. In this study, reduction of water activity and moisture content of the PSPP with the increase in steam pressure at 300, 400 and 500 kPa was due to an increase of the drum surface temperature to 104.67 ± 2.05, 114.67 ± 2.05 and 128.00 ± 1.63 • C, respectively. Similarly, Kakade et al. [26] highlighted that increasing steam pressure amplifies the drum surface temperature and increases moisture evaporation from the feed material and, hence, reduces the product's MC. Valous et al. [28] reported a similar effect on the MC of drum-dried pregelatinized maize starches. Kakade et al. [26] further explained that high DS reduces the retention time of feed on the heated drum surface which increases the MC of the final product. Dao [25] also reported that the moisture content of the drum-dried pumpkin powder decreased with the reduction of DS but conversely increased with steam pressure. Table 2 summarizes the effect of CA, SP and DS on different color attributes (L*, a*, b*, hue and chroma) of the PSPP. The changes in SP and DS had no significant (p > 0.05) effect on the L*, b* and hue values, but had a significant (p < 0.05) effect on CA. The addition of CA had a significant positive effect on the a*, b*, hue and chroma values and a significant (p < 0.05) negative effect on L* (Table 4). Fan et al. [18] found a significant effect of pH on the hue and chroma values in their study on the color stability of anthocyanins in fermented purple sweet potatoes, which was in agreement with our findings.

Color
Concurrently, DS had a significant positive effect on the a* and chroma values. Valous et al. [28] stated that DS was negatively related to the drum's surface temperature at increasing DS. The destruction by heat could be minimized if the retention time of PSP puree on a heated drum surface was shorter, hence it would intensify the a* and chroma values. The quadratic effect of CA had a significant positive relationship with L* but a significant negative relationship with the a*, b*, hue and chroma values. On the other hand, the quadratic effect of SP and DS showed a significant (p < 0.05) positive relationship with the L* value of the drum-dried PSPP. Pua et al. [21] also reported that there was no significant effect of SP on the L* and a* values of the drum-dried jackfruit powder.
Color is one of the most important attributes of food that attract consumers towards buying a particular food product. The pH-dependent molecular structure of anthocyanins results in various natural colors due to structural changes in the flavylium cation, colorless carbinol pseudobases, quinoidal bases, and yellow chalcone with pH changes affecting the color stability of anthocyanins [18]. According to He et al. [13], PSP anthocyanins are more stable at low pH as they predominantly exist as flavylium cations. Acylated peonidin and cyanidins are the predominant anthocyanins in PSP and these acylated structures have a beneficial effect on color stability in acidic conditions, possibly accounting for a more intense reddish-to-pink color of the PSPP when the CA level increases. However, there was a slight color change of PSPP when SP and DS were varied compared to CA indicating that processing variables during drum drying could reduce the degradation of anthocyanins and intensify red color formation which may mask color changes due to Maillard browning reactions and caramelization [25,45]. Figure 4a,b show the color improvement of the drum-dried PSPP with and without CA under optimum SP and DS. There was a significant (p < 0.05) strong negative correlation (r = −0.980) between chroma and pH of the PSPP.

Total Anthocyanin Content
CA had a positive effect on the total anthocyanin content ( Figure 5), whereas S a significant (p < 0.05) negative effect (Tables 3 and 4), ranging from 101.82 to mg/100 g (Table 2). Both quadratic effects of CA and SP had significant (p < 0.05) ne and positive effects on the TAC, respectively, whereas DS had no significant (p > effect on the total anthocyanin content. As shown in Table 3, the reduced model

Total Anthocyanin Content
CA had a positive effect on the total anthocyanin content ( Figure 5), whereas SP had a significant (p < 0.05) negative effect (Tables 3 and 4), ranging from 101.82 to 124.09 mg/100 g (Table 2). Both quadratic effects of CA and SP had significant (p < 0.05) negative and positive effects on the TAC, respectively, whereas DS had no significant (p > 0.05) effect on the total anthocyanin content. As shown in Table 3, the reduced model of the TAC illustrated that the increase in CA had a more positive significant effect on the TAC than other drum drying parameters. This may be due to the acylated structure of anthocyanin under acidic conditions whereby the predominant protonated flavylium cations are more stable to heat and light as a result of intramolecular co-pigmentation. The presence of a more stable pigmented complex of PSP anthocyanins may lead to positive results with a higher CA content [13,46]. Jie Li et al. [12] reported that pH significantly influenced the stability of PSP anthocyanins and a higher anthocyanin content was detected at pH 2 and 3 than at pH 5 and 6. Similarly, Fan et al. [18] discovered that PSP anthocyanins are more stable at pH 2-4 than at the natural pH of PSP i.e., pH 5-6. There was a significant (p < 0.05) strong negative correlation (r = −0.903) between the TAC and pH of the PSPP in the presence of CA. This could be the plausible explanation as to why CA had a positive relationship with the TAC. Steam pressure had a significant negative effect on the TAC due to the higher surface temperature of the drum which degraded the heat-labial anthocyanins present in PSP [47]. This observation is similar to that of Durge et al. [48] who reported that retention of anthocyanins decreased with increase in the temperature of in rice extrudates, emphasizing the susceptibility of anthocyanins to high temperature. These findings were also supported by Hou et al. [49] in their study on black rice anthocyanins, whereby the degradation rate of the total anthocyanin in black rice increased at higher pH and heating temperature. Table 2 shows the effect of CA, SP and DS on the DPPH, FRAP, TPC and TFC of the drum-dried PSPP. The DPPH of the drum-dried PSPP varied from 20.405 to 30.787 μmol TE/g, and a significant (p < 0.05) positive effect of DS and quadratic effect of CA on DPPH Steam pressure had a significant negative effect on the TAC due to the higher surface temperature of the drum which degraded the heat-labial anthocyanins present in PSP [47]. This observation is similar to that of Durge et al. [48] who reported that retention of anthocyanins decreased with increase in the temperature of in rice extrudates, emphasizing the susceptibility of anthocyanins to high temperature. These findings were also supported by Hou et al. [49] in their study on black rice anthocyanins, whereby the degradation rate of the total anthocyanin in black rice increased at higher pH and heating temperature. Table 2 shows the effect of CA, SP and DS on the DPPH, FRAP, TPC and TFC of the drum-dried PSPP. The DPPH of the drum-dried PSPP varied from 20.405 to 30.787 µmol TE/g, and a significant (p < 0.05) positive effect of DS and quadratic effect of CA on DPPH was observed. There was a significant (p < 0.05) negative value for the quadratic effect of SP and interaction of CA with SP. Meanwhile, CA and SP showed no significant (p > 0.05) effect on DPPH of the drum-dried PSPP. Jing Li et al. [10] reported that the thermal degradation of DPPH radical scavenging activity of PSP was not extreme; with only a 5.0% loss in radical scavenging activity observed in PSP anthocyanins in the citric acid-sodium citric buffer after 48 h at 90 • C.

Antioxidant Activity
The FRAP values observed in this study varied from 41.500 to 61.998 µmol TE/g. There was a significant (p < 0.05) positive effect of DS and quadratic effect of CA on the FRAP. The DPPH and FRAP increased with the increased in drum rotation speed, possibly due to the reduction in retention time of the feed on the heated drum surface, thereby causing less degradation of heat-labile anthocyanins and phenolic compounds in PSP. Furthermore, a strong positive (r = 0.838) and significant (p < 0.05) correlation observed between DPPH and FRAP, suggested that the scavenging activity and reducing power of the PSPP were positively related to each other.
There are many studies regarding the potential health benefits of phenolic compounds [50]. The TPC of the drum-dried PSPP ranged from 11.800 to 14.635 mg GAE/g and a significant (p < 0.05) negative effect of SP and interaction of CA with SP was detected. Meanwhile, the DS and the quadratic effect of CA and DS were significantly positive on the TPC of the PSSP. The reduction in the TPC with increasing steam pressure ( Figure 6) was due to the high temperature of the drum surface leading to damage of some heat-labile bioactive and other phenolic compounds. The increase of the TPC with high DS was due to the reduced retention time of the feed on the heated drum surface. Xu et al. [50] reported that increasing temperature is more critical to the heat-labile phenolics because higher thermal energy influences the stability of bioactive compounds. There are many studies regarding the potential health benefits of phenolic compounds [50]. The TPC of the drum-dried PSPP ranged from 11.800 to 14.635 mg GAE/g and a significant (p < 0.05) negative effect of SP and interaction of CA with SP was detected. Meanwhile, the DS and the quadratic effect of CA and DS were significantly positive on the TPC of the PSSP. The reduction in the TPC with increasing steam pressure ( Figure 6) was due to the high temperature of the drum surface leading to damage of some heat-labile bioactive and other phenolic compounds. The increase of the TPC with high DS was due to the reduced retention time of the feed on the heated drum surface. Xu et al. [50] reported that increasing temperature is more critical to the heat-labile phenolics because higher thermal energy influences the stability of bioactive compounds. Phenolic compounds, anthocyanins and β-carotene are mainly responsible for the antioxidant activity of foods; hence, the DPPH radical scavenging activity of PSP is mainly due to anthocyanins and other phenols [45]. Even though the antioxidant activity of food may be affected by the amount of anthocyanins and phenolic compounds, there was no correlation (p > 0.05) between some independent variables in this study. Shih et al. [45] also found no correlation (p > 0.05) between the response variables of the antioxidant capacity and other physicochemical properties of sweet potatoes. Table 2 shows the effects of CA, SP and DS on the WSI and WAC of the drum-dried Phenolic compounds, anthocyanins and β-carotene are mainly responsible for the antioxidant activity of foods; hence, the DPPH radical scavenging activity of PSP is mainly due to anthocyanins and other phenols [45]. Even though the antioxidant activity of food may be affected by the amount of anthocyanins and phenolic compounds, there was no correlation (p > 0.05) between some independent variables in this study. Shih et al. [45] also found no correlation (p > 0.05) between the response variables of the antioxidant capacity and other physicochemical properties of sweet potatoes. Table 2 shows the effects of CA, SP and DS on the WSI and WAC of the drum-dried PSPP. The WSI of the drum-dried PSPP varied from 27.425 to 30.927%, with the WAC ranging between 663.252 and 752.04. There was a significant (p < 0.05) positive effect of CA and DS on the WSI, but the quadratic effect of CA and SP was found to be in significant. Figure 7 depicts the effect of independent variables on the WAC of the PSPP, which was positively and significantly (p < 0.05) affected by DS, the quadratic effect of CA as well as DS and the interaction effect of CA with DS. Shih et al. [45] stated that WSI is related to the degradation of starch during processing and, according to Supprung et al. [27], this is due to the shear forces of the drum and heat causing a reduction in the water absorption capacity of starch. The positive relationship between the WAC and DS observed in this study could possibly be due to the reduced retention time of the feed on the heated drum surface. The WSI was found to be negatively correlated to pH as the addition of CA reduced the pH of the PSPP. Moreover, starch decomposition may increase under acidic conditions, leading to an increase in the WSI with the increase in CA.

Optimisation of Drum Drying Process Parameters for Production of PSPP
Multiple numerical optimization plots were drawn and the data are summarized in Table 5 to determine the optimum processing conditions. For overall desirability (0.8181), the concentration of citric acid, the drum drying variables of steam pressure and drum rotation speed were determined as 0.59%, 499.8 kPa and 3 rpm, respectively. These optimum conditions were predicted to lower the MC (6.791% db) and maximize the chroma

Optimisation of Drum Drying Process Parameters for Production of PSPP
Multiple numerical optimization plots were drawn and the data are summarized in Table 5 to determine the optimum processing conditions. For overall desirability (0.8181), the concentration of citric acid, the drum drying variables of steam pressure and drum rotation speed were determined as 0.59%, 499.8 kPa and 3 rpm, respectively. These optimum conditions were predicted to lower the MC (6.791% db) and maximize the chroma (39.487), a* (39.190), TAC (108.005 mg/100g), DPPH (27.53 µmol TE/g), FRAP (56.319 µmol TE/g), TPC (13.858 mg GAE/g), pH (4.144) and WAC (736.396%).
Drum drying is a heat and mass transfer process which involves high energy utilization and the conditions can be optimized to provide an acceptable high-quality product [21]. In this study, the suggested optimized values for the drum dryer for production of an acceptable high-quality PSPP are high steam pressure and moderate-high drum rotation speed of 499.8 kPa and 3 rpm, respectively.

HPLC Quantification of PSPP Anthocyanins
Since authentic anthocyanin standards were unavailable, the anthocyanin content of the PSPP was calculated as cyanidin and peonidin after acid hydrolysis of the PSPP and the peaks were compared with standard samples. Based on the retention time of the standards, peaks 2 and 3 were identified as cyanidin and peonidin, respectively (Figure 8b). Huang et al. [36] suggested that the elution order of the six major anthocyanidins should be as follows: delphinidin, cyanidin, petunidin, pelargonidin, peonidin and malvidin. However, the sample retention time of peak 1 (Figure 8b) did not match with the standard delphinidin, which confirmed that peak 1 did not belong to the six major anthocyanins. Furthermore, Fan et al. [18] reported that PSP anthocyanins mainly comprised of mono-or diacylated forms of peonidin and cyanidin. In this study, it was found that the anthocyanin content obtained by HPLC was higher than the value obtained using the pH differential method (Table 7); a similar trend was observed by Kang et al. [51] and Lee et al. [52]. Further studies are needed to determine the anthocyanidins content of the PSPP. Furthermore, Fan et al. [18] reported that PSP anthocyanins mainly comprised of monoor diacylated forms of peonidin and cyanidin. In this study, it was found that the anthocyanin content obtained by HPLC was higher than the value obtained using the pH differential method (Table 7); a similar trend was observed by Kang et al. [51] and Lee et al. [52]. Further studies are needed to determine the anthocyanidins content of the PSPP.

Thermal Properties of the Optimized PSP Powder
The impact of optimum drum drying conditions on the thermal properties of the PSPP was analyzed. The results showed that there was no endothermic peak observed in the DSC analysis. This finding confirmed that the drum-dried PSP starch is in an amorphous state, hence the optimum processing conditions resulted the production of an instant (pregelatinized) PSPP [53].

Conclusions
The drum drying conditions for the production of the PSPP were optimized using RSM with the three-factor BBD, namely, citric acid level (0.5-1.5%), steam pressure (300-500 kPa) and drum rotation speed (1-3 rpm). The addition of citric acid positively influenced the color parameters and TAC of the PSPP, whereas the steam pressure and drum rotation speed significantly affected the MC and antioxidant properties, with high steam pressure and low drum rotation speed providing low a w and MC. The optimal processing condition for production of PSPP was achieved with steam pressure of 499.8 kPa and drum rotation speed of 3 rpm and addition of 0.59% citric acid at the overall desirability value of 0.8181 with the acceptable level of physicochemical and functional properties.