IgGM: A Generative Model for Functional Antibody and Nanobody Design 1/ IgGM is a novel generative model designed to facilitate the de novo design of antibodies and nanobodies with functional specificity, focusing on their ability to bind specific antigens. This model uniquely generates both the sequences and structures of antibodies in a single step. 2/ The model consists of three main components: a pre-trained language model that extracts sequence features, a feature learning module to identify relevant features, and a prediction module to generate antibody sequences and predict the antibody-antigen complex structure. 3/ IgGM addresses a major challenge in antibody design by generating novel antibody sequences and their structures without the need for existing experimental antibody-antigen complex structures, making it suitable for scenarios involving new antigens. 4/ The model is capable of designing antibody CDR regions, which are crucial for antigen binding, and can also predict and design entire antibody structures, a significant advancement over existing methods that often require pre-existing structures or templates. 5/ IgGM excels in both antibody and nanobody design tasks, achieving superior performance in multiple design scenarios, such as generating antibodies with high specificity and binding affinity to antigens, outperforming previous methods in terms of sequence and structure fidelity. 6/ Its ability to design antibodies and nanobodies for specific epitopes, including the design of multiple CDR regions simultaneously, provides a versatile solution for therapeutic and diagnostic applications. 7/ Experimental results demonstrate IgGM's impressive accuracy, including higher success rates in docking performance and lower RMSD values compared to existing methods, showcasing its potential in the rapidly growing field of AI-driven drug design. 💻Code: github.com/TencentAI4S/IgGM 📜Paper: biorxiv.org/content/10.1101/… #AntibodyDesign #NanobodyDesign #GenerativeModels #MachineLearning #AIinHealthcare #DrugDiscovery #Bioinformatics #ComputationalBiology #DeepLearning #AIforMedicine #ProteinDesign