32nd IEEE Conference on Signal Processing and Communications Applications, SIU 2024, Mersin, Türkiye, 15 - 18 Mayıs 2024
Assessing drug-likeness in chemical compounds is a critical step in drug discovery and development. Traditional approaches often rely on Lipinski's Rules to evaluate the drug-likeness of compounds; however, recent advances in machine learning offer alternative methods. This study presents a comprehensive comparison of various machine learning methods for predicting drug similarity. Classification based on Lipinski Descriptors was performed to prepare the dataset and train the models. The research includes the comparison of many machine learning algorithms in terms of performance and accuracy in the mentioned context. The performance of these methods was evaluated in terms of accuracy, sensitivity, and specificity, considering their ability to distinguish between drug-like and non-drug-like compounds. The dataset includes a wide range of chemical compounds and has been compiled in the chembl database. The findings of this study shed light on the effectiveness of machine learning approaches in predicting drug similarity without relying on predefined rules. Additionally, information on the strengths and limitations of different algorithms in processing Lipinski Descriptors is discussed. This work contributes to ongoing work on improving computational methods for drug discovery and may guide future similar work by highlighting the potential of machine learning in optimizing the early stages of drug development.