This review discusses advances in de novo protein design, which aims to create new proteins from scratch by specifying a desired function and designing a structure that can achieve it. The central dogma of de novo design involves mapping functional goals to structural motifs, controlling and designing protein structure, and sampling sequences to achieve the desired function. Recent progress in deep learning has enabled more accurate and efficient protein design, allowing for the creation of functional proteins from scratch rather than modifying existing ones. The review highlights the importance of understanding protein folding and function, and how computational methods can capture this knowledge to enable protein design. It also discusses the challenges of designing proteins with complex functions, such as antibodies and enzymes, and the need for new approaches that consider conformational dynamics and heterogeneity. The review emphasizes the potential of deep learning-based methods for protein design, including generative models that can produce de novo proteins with high accuracy and efficiency. It also addresses the limitations of current methods, such as the need for more data and the challenge of predicting protein function. The review concludes that while significant progress has been made in de novo protein design, there are still many challenges to overcome, including the need for more comprehensive models that can account for the complexity of protein function and structure. The review also highlights the importance of continued research in this area to improve the accuracy and efficiency of protein design methods.This review discusses advances in de novo protein design, which aims to create new proteins from scratch by specifying a desired function and designing a structure that can achieve it. The central dogma of de novo design involves mapping functional goals to structural motifs, controlling and designing protein structure, and sampling sequences to achieve the desired function. Recent progress in deep learning has enabled more accurate and efficient protein design, allowing for the creation of functional proteins from scratch rather than modifying existing ones. The review highlights the importance of understanding protein folding and function, and how computational methods can capture this knowledge to enable protein design. It also discusses the challenges of designing proteins with complex functions, such as antibodies and enzymes, and the need for new approaches that consider conformational dynamics and heterogeneity. The review emphasizes the potential of deep learning-based methods for protein design, including generative models that can produce de novo proteins with high accuracy and efficiency. It also addresses the limitations of current methods, such as the need for more data and the challenge of predicting protein function. The review concludes that while significant progress has been made in de novo protein design, there are still many challenges to overcome, including the need for more comprehensive models that can account for the complexity of protein function and structure. The review also highlights the importance of continued research in this area to improve the accuracy and efficiency of protein design methods.