This review focuses on recent advances in structure-based *de novo* drug design, a subset of computer-aided drug design (CADD) that aims to generate novel molecules with pharmacological properties from scratch. The review covers various sampling methods, including conventional fragment-based methods, evolutionary algorithms, Monte Carlo Metropolis methods, and deep generative models. It highlights the challenges and solutions in synthetic accessibility (SA) and the importance of benchmarking strategies to validate proposed frameworks. The review also discusses the role of scoring functions in evaluating ligand properties and the limitations of current methods, such as the need for more accurate scoring functions and multiobjective optimization to address SA concerns. Finally, it emphasizes the importance of standardized benchmarking workflows and public benchmarking exercises to advance the field of structure-based *de novo* drug design.This review focuses on recent advances in structure-based *de novo* drug design, a subset of computer-aided drug design (CADD) that aims to generate novel molecules with pharmacological properties from scratch. The review covers various sampling methods, including conventional fragment-based methods, evolutionary algorithms, Monte Carlo Metropolis methods, and deep generative models. It highlights the challenges and solutions in synthetic accessibility (SA) and the importance of benchmarking strategies to validate proposed frameworks. The review also discusses the role of scoring functions in evaluating ligand properties and the limitations of current methods, such as the need for more accurate scoring functions and multiobjective optimization to address SA concerns. Finally, it emphasizes the importance of standardized benchmarking workflows and public benchmarking exercises to advance the field of structure-based *de novo* drug design.