Protein-ligand modeling is a cornerstone of modern drug discovery, facilitating the identification and optimization of bioactive compounds that modulate protein function. Computational approaches provide cost-effective and scalable strategies for exploring the growing chemical and biological spaces, accelerating early-stage drug development. Advances in both physics-based methods and data-driven machine learning techniques have expanded the range and accuracy of tools available for modeling protein-ligand interactions. This review provides a current and concise view of key methodologies in protein-ligand modeling, including binding site prediction and the generation and evaluation of target-bound ligand conformations. It also discusses state-of-the-art machine learning approaches that are reshaping how these tasks are performed and enhancing the accuracy of binding site, binding pose, and binding affinity predictions.