Protein structure prediction uses computational methods to determine a protein's 3D structure from its amino acid sequence. Modern AI systems use neural networks that incorporate knowledge of physical and biological constraints through multi-sequence alignments and attention mechanisms.
For more technical details, see our research documentation.
Soy3D Atlas is a comprehensive resource that provides access to protein structure data and analysis tools. It offers predicted 3D structures, visualization capabilities, and various bioinformatics tools for researchers.
The platform aims to accelerate scientific research by making structural data and analysis tools accessible to the global scientific community.
Prediction accuracy varies depending on the protein and method used. Modern AI-based approaches have achieved significant improvements in recent years, with some achieving average Global Distance Test (GDT) scores above 90 for certain protein categories.
Each prediction typically includes confidence scores (such as pLDDT), which indicate how confident the model is about the predicted position of each amino acid residue.
You can search the database by gene name, UniProt ID, protein sequence, or keyword. For each protein, you can view the predicted 3D structure, download coordinate files (PDB format), and access related information such as confidence scores and structural analysis.
The platform also provides various bioinformatics tools and links to external resources such as UniProt, PDB, and other databases.
Yes, structural data can be downloaded from our download page. The data is provided in standard formats such as PDB files, along with associated metadata.
We also provide programmatic access to the database through our API.