Chemicals
The chemicals used in this study were all of analytical grade and all the substrates viz. wheat bran, rice husk, soybean meal and black gram husk were purchased from local market at Teliarganj, Allahabad-211004, India.
Maintenance of culture and inoculum preparation
For the present work, Rhizopus oryzae (SN5)/NCIM-1447strain isolated from compost soil obtained from Ranikhet, Uttrakhand, India was utilized. This strain was characterised from NCIM, Pune India and deposited. Strain Seq286_NC111119 showed closest homology with Rhizopus sp. (closer to oryzae). This strain was maintained on Potato Dextrose Agar (PDA) slants at 4oC. For preparation of inoculum, 25mL of sterile distilled water was added to five days old slant and aseptically scraped with an inoculating loop. This suspension of spores was used as inoculum for the preparation of induced culture.
Preparation of induced culture
Medium for preparation of induced culture contained 1% casein, 0.25% glucose in modified Czapek-Dox broth. This medium was autoclaved and then inoculated with 2% of the spore suspension containing approximately 3.2 x 106 spores/ml. The medium was incubated at 28oC for 6 days. After the complete growth of the organism, the flask was placed on a shaker to make a homogenous suspension of spores.
Fermentation set up
The fermentation was carried out in 500mL Erlenmeyer flasks, each carrying 5gm of different substrates viz. wheat bran, soybean meal, a mixture of wheat bran and soybean meal in ratio 4:1, black-gram husk and rice husk, respectively. The substrates were soaked in 10mL of Czapek-Dox salt solution with pH 6 and autoclaved. This preparation was then inoculated with 1mL of the induced spore suspension containing about 4.3 x 107 spores and kept in incubator at 28oC. Fermentation was carried out in different experimental conditions changing one parameter at a time.
Extraction of crude enzyme
The fermented substrate was soaked in 50mL of 0.1 M phosphate buffer and kept on a shaker at 120 rpm for 2 hours. The substrate was filtered through muslin cloth and the extract was centrifuged at 10,000 rpm at 4oC for 10 minutes for removal of spores and other insoluble debris. The supernatant or crude enzyme extract was stored at 4oC for measurement of protease activity.
Protease assay
Protease activity was measured through modified caseinolytic method (Walter 1984). 0.65mL of glycine-NaOH buffer with pH 8 was incubated with 0.05mL of enzyme solution for 10 minutes. 2mL of 2% casein solution was then added and the reaction mixture was incubated at 37oC for 20 min. The reaction was ceased by adding 0.1 mL of 1 M HCl and the non-hydrolysed casein was precipitated by 5 mL of 5% Trichloroacetic acid. The precipitated casein was removed through centrifugation at 10,000 rpm for 5 minutes. The protein concentration was measured in 0.5 mL of the supernatant through Lowry method using BSA as standard (Lowry et al. 1951). One unit of enzyme activity is defined as the amount of enzyme that liberates peptide fragments equivalent to 1 mg of BSA per minute under the assay conditions (Patil and Shastri, 1981).
Effect of Process parameters on enzyme production: carbon sources, incubation time, pH, temperature,
Various process parameters such as temperature, carbon sources, incubation time, pH, carbon and nitrogen sources, were studied to monitor their effect on alkaline protease production in Solid-State fermentation through one at a time first and then with EVOP factorial design technique.
To screen carbon source, 5 g each of different agro-industrial wastes such as wheat bran, rice husk, black-gram husk, soybean meal and a mixture of wheat bran and soybean meal in ratio 4:1 (in g) were used as substrate in different combinations for fermentation. The fermentation flasks were incubated for 24, 48, 72, 96, 120 and 144 hours and protease activity was measured according to the procedure described earlier. The ratio of Wheat Bran and Soybean Meal was changed as follows- 4.5:0.5, 4:1, 3:2, 2:3 (in g) and fermentation was done as described before. In order to optimum nitrogen supplement to achieve max yield czapek-dox salt solution was supplemented with sodium nitrate (1%), ammonium nitrate (1%), peptone (1%), yeast extract (1%) and potassium nitrate (1%) to study the impact of different nitrogen sources on alkaline protease production.
Physical parameters like initial pH and temperature were optimized by maintaining the pH of Czapek-Dox broth using 1N HCl/1N NaOH as 4, 5, 6, 7 and 8 before autoclaving media and temperature was kept 28, 30, 35 and 37oC by keeping the fermentation flasks in different incubators pre-set at respective temperature.
Process parameters optimization through EVOP factorial Design technique
The EVOP methodology was applied to determine the optimum levels of three parameters (ratio of wheat bran to soybean meal, pH and temperature) in different experiments. Firstly, the control experimental conditions (A1 and A6) were selected based on the results obtained from one factor at a time optimization. Then, new experimental conditions with higher and lower search levels of parameters were selected. The experimental design for three variables system is shown in Table 1. Thirdly, fermentation was carried out as described earlier at higher and lower search levels and all the experiments were repeated for two cycles. The yield of alkaline protease in cycle I and cycle II was measured following the procedure mentioned earlier. The differences in protease yield in cycle I and cycle II were calculated along with the averages in order to estimate the effects and error limits. The magnitude of the effects, error limits and changes in the main effect were inspected as per the decision making procedure to arrive at the optimum.
Table 1: Experimental Design for three variables system
Parameters
|
A1
|
A2
|
A3
|
A4
|
A5
|
A6
|
A7
|
A8
|
A9
|
A10
|
Temperature/ oC
|
0
|
-
|
-
|
+
|
+
|
0
|
+
|
-
|
+
|
-
|
pH
|
0
|
-
|
+
|
-
|
+
|
0
|
+
|
-
|
-
|
+
|
WB +SM ratio
|
0
|
-
|
+
|
+
|
-
|
0
|
+
|
+
|
-
|
-
|
Response
|
a1
|
a2
|
a3
|
a4
|
a5
|
a6
|
a7
|
a8
|
a9
|
a10
|
--------------------> Block I<-------------------------------- ----------------------->Block II <------------------
Decision making
Once all the calculations for effects and error limits are completed, it is extremely necessary to determine whether any changes in the control level will help to enhance the objective function and if the changes are required then the next step is to identify the desired direction of change. For this study, magnitudes of the effects were compared to those of the error limits. If all or any of the effects are higher than the error limits, alterations in the experimental conditions may yield better results.
Optimization of process parameters through ANN Method
Data selected for optimization
Data used in ANN was output of Evop-Factorial design technique and one factor at a time i.e., pH, incubation temperature and different percentage of two different substrate combination of Wheat bran(WB)and Soybean mil(SM) [Table 2].
Table- 2 shows the data selected to optimize the protease activity.
Temperature
|
WB% in (WB+SM)
|
pH
|
Protease Activity
|
28
|
80
|
6
|
398.03
|
26
|
70
|
4
|
144.45
|
26
|
90
|
8
|
49.38
|
30
|
90
|
4
|
93.74
|
30
|
70
|
8
|
144.1
|
28
|
80
|
6
|
398.47
|
30
|
90
|
8
|
195.13
|
26
|
90
|
4
|
114.85
|
30
|
70
|
4
|
106.97
|
The data selected for optimization was used to fit a nonlinear regression:
Protease activity= -2299.6275+7.896875*x(1)-0.591625*x(2)+839.270625*x(3)-69.7646875*x(3)^2
Where :x(1)=temperature value ,x(2)=WB percentage in the WB and SM substrate mixture ,
x(3)=pH value.
Optimization methodology: This analysis was carried out in MATLAB, Windows 8 operating system, Intel R CPU 2.53 GHz, 8.00 GB of RAM. The input data and target data was the data collected from the Table 2, which was fed into the designed neural network for training. The speed of this process varies according to the specifications of the system. There are different types of neural network available for training and optimization. However, we had chosen to solve this particular problem using feed forward network and trained by back propagation algorithm. The network flow is only in one direction. There was no feedback because we wanted to create a simple network for this problem. This network used supervised learning so input and output data were fed to the network. In case of unsupervised learning, only input has given and linear equations were built such that maximum correlation coefficient was achieved. Signals flow forward and errors are propagated backward to ensure that errors were reduced and R square value increased closer to 1. Random weights were initialized and changed in each run of each iteration to made change in learning process. The objective was to minimize the errors. The different parameters considered in the neural network included training functions, performance measures, number of layers, number of hidden layers and neurons, and transfer function.
Network Architecture and Training of MFNN: A three layer (3-10-1) MFNN was designed using MATLAB: 1 input layer (3 neurons),1 hidden layer(10),1 output layer(1 neuron). Layers were kept fully connected means each neuron in next layer has been connected to each neuron in previous layer. The sigmoidal function was used as transfer function of neurons in hidden layer and linear function was used as transfer function of output layer. Sigmoidal function gives more accuracy with backpropagation algorithm. ‘Trainlm’ was used for training the MFNN because it is a fast and stable algorithm[M63]. The network was trained for 64 times to find best fit to the data. The training results of The R square value was found 0.9967 for overall, 0.99909 for Training, 0.9994 for validation and 0.988 for text, which is very close to one and acceptable for good fitting. Performance of MFNN is shown in Figure (1a) and Figure (1b).
Calculating network output: The trained neural network was used to generate protease activity at various input values of temperature, WB%, and pH value setting constraints.
(1) 26<=temperature<= 30 at increment of 1 oC
(2) 70<=WB%<=90 at increment of 5
(3) 4<=pH<=8 at increment of 0.2 .
Network output was calculated at 1850 possible combinations of inputs to achieve our goal of optimization of protease activity. The network output has displayed in Figure (2a) and (2b). Individual combination of temperature, pH and WB % was given an iteration number which was plotted in x-axis and the corresponding protease activity of network output was plotted in y-axis as shown in the Figure (2a).
The maximum protease activity was between 1000th to 1100th iterations. The network response of between these values was further plotted in the Figure (2b). It was found again that maximum protease activity was in between 1020th to 1030th iteration. The network response for these values is shown in the Table 4.
Table 2 Experimental conditions and results of experimental setup I for protease production
Parameters
|
A1
|
A2
|
A3
|
A4
|
A5
|
A6
|
A7
|
A8
|
A9
|
A10
|
Temperature/ oC
|
28
|
26
|
26
|
30
|
30
|
28
|
30
|
26
|
30
|
26
|
pH
|
6
|
4
|
8
|
4
|
8
|
6
|
8
|
4
|
4
|
8
|
WB +SM ratio (in g)
|
4:1
|
3.5:1.5
|
4.5:0.5
|
4.5:0.5
|
3.5:1.5
|
4:1
|
4.5:0.5
|
4.5:0.5
|
3.5:1.5
|
3.5:1.5
|
Protease Activity (U/gds)
(Cycle 1)
|
406.17
|
158.97
|
48.63
|
88.80
|
153.34
|
412.79
|
189.32
|
107.69
|
110.26
|
99.30
|
Protease Activity (U/gds)
(Cycle 2)
|
389.89
|
129.93
|
50.12
|
98.69
|
134.86
|
384.16
|
200.94
|
122.02
|
103.69
|
110.53
|
Differences
|
16.28
|
29.04
|
-1.49
|
-9.89
|
18.48
|
28.63
|
-11.62
|
-14.33
|
6.57
|
-11.23
|
Average
|
398.03
|
144.45
|
49.38
|
93.74
|
144.10
|
398.47
|
195.13
|
114.85
|
106.97
|
104.91
|
Table 3 Calculations of effects and error limits for cycles I and II for amylase and protease
|
Protease (Cycle 1)
|
Protease (Cycle 2)
|
Effects
|
Effect of Temperature (T)
|
-26.05
|
-7.55
|
Effect of pH (P)
|
11.70
|
7.11
|
Effect of WB +SM ratio (R)
|
30.50
|
40.56
|
Effect of TP
|
13.22
|
13.81
|
Effect of TR
|
34.59
|
28.93
|
Effect of PR
|
61.39
|
50.43
|
Change in Mean Effect
|
-231.94
|
-214.54
|
Standard Deviation
|
100.65
|
93.03
|
Error limits
|
For average
|
±142.32
|
±131.54
|
For effects
|
±101.04
|
±93.40
|
For change in mean
|
±89.68
|
±82.89
|
Table 4: Network response of maximum protease activity between 1020th to 1030th iteration
Iterations
|
1021
|
1022
|
1023
|
1024
|
1025
|
1026
|
1027
|
1028
|
1029
|
1030
|
Temperature
|
29
|
29
|
29
|
29
|
29
|
29
|
30
|
30
|
30
|
30
|
percentage of WB
|
77.5
|
80
|
82.5
|
85
|
87.5
|
90
|
70
|
72.5
|
75
|
77.5
|
pH
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
6.2
|
protease activity
U/gds
|
413.82
|
411.02
|
412.42
|
416.95
|
415.1
|
409.35
|
422.66
|
421.14
|
420.72
|
420.41
|