Physicochemical Properties of Some Drugs by Using Machine Learning

Abstract

This study explores the integration of machine learning and multi-criteria decision-making techniques for the characterization and evaluation of potential therapeutics targeting Duchenne Muscular Dystrophy . A dataset comprising the physicochemical properties of twenty anti- Duchenne muscular dystrophy drug candidates is analyzed using topological indices de- rived from molecular graphs. These indices, computed via Python-based algorithms, served as input features for machine learning models. Artificial Neural Networks and Random Forests are employed to predict key physicochemical properties based on the structural de- scriptors. The results demonstrate that topological indices are effective predictors, enabling accurate and consistent performance across both models. To prioritize the drug candidates, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is applied as an multi-criteria decision-making tool, leveraging the model predictions to rank compounds based on multiple beneficial attributes. The proposed framework highlights the synergy be- tween graph-theoretical descriptors, machine learning, and multi-criteria decision-making, offering a robust in-silico approach for early-stage drug evaluation. This methodology pro- vides a systematic and objective strategy for ranking therapeutic candidates, particularly valuable in the context of rare diseases such as duchenne muscular dystrophy. Key words: Machine learning; Molecular Graph; QSPR Analysis; Topological indices; Ar- tificial Neural Networks; Random Forest; TOPSIS; Python Algorithm; Anti-DMD drugs.