A5081693954- Protein Structure and Dynamics
- Monoclonal and Polyclonal Antibodies Research
- Computational Drug Discovery Methods
- Machine Learning in Bioinformatics
- Bioinformatics and Genomic Networks
- Genomics and Phylogenetic Studies
- vaccines and immunoinformatics approaches
- RNA and protein synthesis mechanisms
- Glycosylation and Glycoproteins Research
- Advanced Biosensing Techniques and Applications
- Innovative Microfluidic and Catalytic Techniques Innovation
- Cell Image Analysis Techniques
- Protein purification and stability
- Neural Networks and Applications
- Invertebrate Immune Response Mechanisms
- Genetics, Bioinformatics, and Biomedical Research
- Enzyme Structure and Function
- Microbial Metabolic Engineering and Bioproduction
- Microfluidic and Capillary Electrophoresis Applications
- Nanofabrication and Lithography Techniques
- Graph Theory and Algorithms
- Machine Learning in Materials Science
- Industrial Technology and Control Systems
Tencent (China)
2022-2025
Institute of Computing Technology
2020-2024
Chinese Academy of Sciences
2020-2024
Toyota Technological Institute at Chicago
2020-2022
University of Chinese Academy of Sciences
2020-2022
Protein structure prediction has been greatly improved by deep learning in the past few years. However, most successful methods rely on multiple sequence alignment (MSA) of homologs protein under prediction. In nature, a folds absence its and thus, MSA-free method is desired. Here, we develop single-sequence-based RaptorX-Single integrating several language models generation module then study advantage over MSA-based methods. Our experimental results indicate that addition to running much...
Abstract Accurate prediction of antibody-antigen complex structures holds significant potential for advancing biomedical research and the design therapeutic antibodies. Currently, structure protein monomers has achieved considerable success, promising progress been made in extending this achievement to complexes. However, despite these advancements, fast accurate remains a challenging unresolved issue. Existing end-to-end methods, which rely on homology templates, exhibit sub-optimal...
Geometric graph is a special kind of with geometric features, which vital to model many scientific problems. Unlike generic graphs, graphs often exhibit physical symmetries translations, rotations, and reflections, making them ineffectively processed by current Graph Neural Networks (GNNs). To tackle this issue, researchers proposed variety equipped invariant/equivariant properties better characterize the geometry topology graphs. Given progress in field, it imperative conduct comprehensive...
Motivation Protein structure prediction has been greatly improved by deep learning, but most efforts are devoted to template-free modeling. But very few learning methods developed for TBM (template-based modeling), a popular technique protein prediction. studied extensively in the past, its accuracy is not satisfactory when highly similar templates available. Results This paper presents new method NDThreader (New Deep-learning Threader) address challenges of TBM. first employs DRNF (deep...
Abstract Accurate prediction of antibody structures is critical in analyzing the function antibodies, thus enabling rational design antibodies. However, existing structure methods often only formulate backbone atoms and rely on additional tools for side-chain conformation prediction. In this work, we propose a fully end-to-end architecture simultaneous conformations, namely tFold-Ab. Pre-trained language models are adopted fast by avoiding time-consuming search sequence homologs. The model...
Abstract Protein structure prediction has been greatly improved by deep learning in the past few years. However, most successful methods rely on multiple sequence alignment (MSA) of homologs protein under prediction. In nature a folds absence its and thus, MSA-free method is desired. Here we develop single sequence-based RaptorX-Single integrating several language models generation module then study advantage over MSA-based methods. Our experimental results indicate that addition to running...
Abstract Alpha-beta T cell receptor ( αβ TCR) recognition of peptide-major histocompatibility complexes (pMHCs) is a cornerstone the adaptive immune system. Fast and accurate modeling TCR-pMHC structures crucial for understanding TCR pMHCs at molecular level, which essential development TCR-based therapeutics vaccines. Despite significant interest, this challenge remains unresolved due to diversity interactions limited structural data. Here, we present tFold-TCR, high-throughput, end-to-end...
The development of therapeutic antibodies heavily relies on accurate predictions how antigens will interact with antibodies. Existing computational methods in antibody design often overlook crucial conformational changes that undergo during the binding process, significantly impacting reliability resulting To bridge this gap, we introduce dyAb, a flexible framework incorporates AlphaFold2-driven to model pre-binding antigen structures and specifically addresses dynamic nature conformation...
Single-domain antibodies (sdAbs) have emerged as powerful therapeutic agents due to their small size, high stability, and superior tissue penetration. However, unlike conventional monoclonal (mAbs), sdAbs lack an Fc domain, limiting functional versatility manufacturability. To address this challenge, we developed TFDesign-sdAb, a deep learning-based generative-ranking framework that enables rational engineering of with tailored functionalities. Our integrates structure-aware diffusion model...
Protein structure prediction has been greatly improved by deep learning, but the contribution of different information is yet to be fully understood. This article studies impacts two kinds for prediction: template and multiple sequence alignment (MSA) embedding. Templates have used some methods before, such as AlphaFold2, RoseTTAFold RaptorX. AlphaFold2 RosetTTAFold only templates detected HHsearch, which may not perform very well on targets. In addition, embedding generated pre-trained...
Protein representation learning plays a crucial role in understanding the structure and function of proteins, which are essential biomolecules involved various biological processes. In recent years, deep has emerged as powerful tool for protein modeling due to its ability learn complex patterns representations from large-scale data. This comprehensive survey aims provide an overview advances techniques applied science. The begins by introducing developments based models emphasizes importance...
Immunoglobulins are crucial proteins produced by the immune system to identify and bind foreign substances, playing an essential role in shielding organisms from infections diseases. Designing specific antibodies opens new pathways for disease treatment. With rise of deep learning, AI-driven drug design has become possible, leading several methods antibody design. However, many these approaches require additional conditions that differ real-world scenarios, making it challenging incorporate...
Humanization is a critical process for designing efficiently specific antibodies and nanobodies prior to clinical trials. Developing widely recognized deep learning techniques or frameworks humanizing conventional presents valuable yet challenging task. Inspired by the effectiveness of diffusion models across various applications, we introduce HuDiff, an adaptive approach from scratch, referred as HuDiff-Ab HuDiff-Nb, respectively. This begins humanization exclusively with...
Abstract Motivation TBM (template-based modeling) is a popular method for protein structure prediction. When very good templates are not available, it challenging to identify the best templates, build accurate sequence-template alignments and construct 3D models from alignments. Results This paper presents new NDThreader (New Deep-learning Threader) address challenges of TBM. DNThreader first employs DRNF (deep convolutional residual neural fields), which an integration deep ResNet...
Abstract In recent years, there has been a growing interest in using deep learning models for protein-ligand docking and affinity prediction, both vital structure based drug design. However, many of these overlook the intricate modeling interactions between ligand protein atoms, thereby constraining their capabilities generalization interpretability. this paper, we introduce \textsc{Interformer}, unified model built upon Graph-Transformer architecture, which specially crafted to capture...