A comparative study of multi-objective methods and algorithms for optimizing emulgels consistency and drug diffusion

The development of effective drug delivery systems is a challenging task that requires a thorough understanding of various destabilization phenomena and the trade-offs between physical and therapeutic properties. In this study, a multi-objective optimization problem was addressed, which involved a conflicting relationship between the consistency of emulgels and their drug diffusion properties. As there is no perfect solution, a compromised solution was sought using thirteen methods and algorithms that are coupled with different modeling techniques, including polynomial RSM and non-linear ML algorithms (ANN and SVR). Sisko's consistency coefficient and the steady-state drug diffusion parameters were modeled using ANN and SVR algorithms. The SVR model outperformed both ANN and RSM based on R2, RMSE, and MAE metrics. Although this performance, it was found that the ANN model was more effective in defining problems for the evolutionary algorithms. Furthermore, AGE-MOEA-II and SMPSO emerged as the top performers in terms of diversity, distribution, and proximity criteria. In particular, evolutionary mechanisms such as particle swarm optimization and adaptive geometry estimation stand out for their excellent performance. The multi-objective optimization yielded a compromised formulation that includes internal almond oil content, mixed surfactants (Span 60/Tween 80) with HLBreq: 11.8, and a carbopol percentage in the aqueous phase of (32.11/5.69/0.96). This resulted in a consistency of 92.29 Pasn and drug diffusion of 182.28 μg.cm−2.h−1. In conclusion, the present study successfully proposes an alternative and more advanced procedure based on ML-EAs to replace the RSM for the optimization of emulgels and other complex problems in drug delivery systems. [Display omitted]