Application of artificial neural networks (ANNs) and genetic programming (GP) for prediction of drug release from solid lipid matrices

From Heuristic to Mathematical Modeling of Drugs Dissolution Profiles: Application of Artificial Neural Networks and Genetic Programming

“…The above equations are even simpler than previously reported ( 10 ) [ 27 ]. This is the result of more thorough investigation of the GP stopping conditions and also longer run times extended up to 120 hours per CPU core where Q is the amount of drug released (%) in time t , d is the extrudate diameter, and c 1–5 are adjustable parameters.…”

“…It was noted that mathematical equations are comparable to the ANNs in their predictive performance. Another performance verification of ( 7 ) was based on the mechanistic modeling presented by Güres et al [ 27 ], where similarity factors ( f 2 ) were computed for the simulated and observed data ( Table 4 ).…”

“…The simplest model found in the indirect modeling procedure was presented in ( 9 ). It contains only 3 adjustable parameters ( c 1 , c 2 , and c 3 ); thus it is not much more complex than the original Weibull equation and certainly less complex than previously reported ( 10 ) [ 27 ]. The direct GP run was meant to be the control for the indirect mode and yet, unexpectedly, it produced a very compact equation ( 7 ) with good predictive abilities ( Table 3 versus Table 2 ) and again with only three adjustable parameters.…”

“…This was clearly beneficial to the less robust ( 9 ), which achieved predictive abilities at the level of ANNs only for this type of the data preprocessing ( Table 3 ). Equation ( 7 ) was also compared in its predictive abilities to the previously derived mechanistic equation [ 27 ] by means of similarity factor ( f 2 ) computed for the modeled versus observed results ( Table 4 ). The results of this comparison favor ( 7 ) over mechanistic model and are in the same time in accordance with the conclusions presented by Güres et al [ 27 ], namely, the low importance of the extrudates length parameter ( L ).…”

“…Equation ( 7 ) was also compared in its predictive abilities to the previously derived mechanistic equation [ 27 ] by means of similarity factor ( f 2 ) computed for the modeled versus observed results ( Table 4 ). The results of this comparison favor ( 7 ) over mechanistic model and are in the same time in accordance with the conclusions presented by Güres et al [ 27 ], namely, the low importance of the extrudates length parameter ( L ). Another confirmation of ( 7 ) robustness was delivered by comparison of the values of its parameters fitted during the 5-fold cross-validation.…”