Computational Prioritization of Bioactive Molecules for Therapeutic and Nutritional Design Using ML–MCDM Approaches

Abstract

Parenteral Nutrition (PN) solutions are specialized intravenous formulations that provide vital nutrients (amino acids) to patients who are unable to meet their nutritional needs orally or enterally. This study emphasizes the significance of acyclic amino acids, which are major components of Parenteral Nutrition (PN) solutions, as they are linear in structure and have effective metabolic applications in clinical nutrition. The present study combines machine learning and multi-criteria decision-making techniques (MCDM), and this combination of approaches is used to rank acyclic amino acids using entropy-based topological indices. This framework is based on regression techniques such as linear, polynomial, LASSO, and Random Forest regression and uses entropy measures of Gourava indices as features. To rank acyclic amino acids, machine learning tools are employed in parallel with MCDM tools (VIKOR and EDAS). The outcomes linked to rankings are tabulated and visualized by means of bar plots and heatmaps to ensure interpretability and transparency. Graphical comparisons and correlation analysis were used to confirm that the ML and MCDM ranks were consistent. Furthermore, the computational rankings are compared to three commercially available PN solutions (Aminoven, TrophAmine, and ProSol) to ensure their clinical relevance. This integration enabled us to compare amino acid priority using ML-MCDM techniques to real-world formulations utilized in parenteral nutrition.