Bioprocess Modeling for the Prediction of Therapeutic Enzyme L- Asparaginase Activity in Solid State Fermentation Using Multiple Linear Regression and Ann

Objective: L-asparaginase is an enzyme of industrial as well as therapeutic importance. The capabilities of bioprocess modeling of L-Asparaginase activity produced from Aspergillus niger by solid state fermentation (SSF) were explored here. Methods: Regression modeling (RM) and Artificial Neural Network (ANN) techniques were applied on input process parameter, which includes solid substrate, temperature, moisture percentage, particle size, cooking time to optimize L-Asparaginase enzyme activity in SSF. Results: The L-asparaginase activity were obtained 38.918 (U/gds) and 38.714 (U/gds) with the optimum input parameters (xx1= Glycine max, xx2=30 (°C), x3=6.5, xx4=70 (%),xx5=1180(μ), xx6=30 min) by ANN, and (xx1= 3, xx2=30 (°C), xx3=6.5, xx4=70 (%), xx5=1305(μ), xx6=30 min) by RM respectively. The goodness of fit of the model was determined in terms of R2. The value of R2 obtained by ANN after training and validation and over all data was 0.996, 0.989 and 0.981, whereas the value of R2 obtained with linear, quadratic and full regression models was 0.501, 0.910 and 0.914 respectively. Conclusion: This hybrid ANN/RM effectively identifies the significant process parameters and optimum production of L-asparaginase in the given larger set of conditions and able to reduce the number of experiments. Optimization by these modeling methods predicts the good activity of the enzyme and indicating its suitability and applicability for bioprocess modeling.