This chapter presents a comprehensive review of in silico techniques for the prediction of solubility, permeability, and metabolism—critical determinants of drug candidates’ pharmacokinetic and safety profiles during drug discovery and development. The introduction highlights the increasing role of computational models, including QSAR/QSPR, machine learning, and physiologically based approaches, in accelerating early-stage assessment and rational design of new molecules, thus minimizing experimental costs and attrition. Key theoretical foundations of solubility—such as thermodynamic and kinetic considerations and the role of molecular descriptors—are dissected, underlining how advanced statistical and AI-driven models improve prediction accuracy, but also face challenges due to dataset limitations and real-world complexity. The chapter details computational tools for solubility, from machine learning platforms (ADMET Predictor, SwissADME) to hybrid thermodynamic and quantum approaches, and explains QSPR and deep learning strategies for property extrapolation. Mechanistic and predictive models for permeability, the use of empirical rules (Lipinski’s Rule of Five), and physiologically based pharmacokinetic (PBPK) modeling are discussed in relation to their relevance for drug absorption and bioavailability screening. For metabolism, it describes computational prediction of enzyme-specific pathways, with an overview of rule-based, docking, and AI methods targeting major metabolic enzymes (CYP450, UGT). The review concludes by evaluating integrated ADME platforms, practical case studies, existing model limitations, and future perspectives on AI, multi-omics data, and model integration—emphasizing their capacity to transform predictive pharmacology and reduce preclinical testing burdens in modern drug development.

By Mani Sharma, Mohini Rawat, Jyoti Sharma, Rama Satya Sri Kotipall, Mrynal Chamoli, Dibyalochan Mohanty