AI Driven Structure Modeling & Validation
Accelerate cryo-EM structure analysis with our AI-powered software suite. Build protein structures with DeepMainmast, model DNA and RNA with CryoREAD, and assess local model quality with DAQ-Score.
All programs are available for commercial use, with flexible licensing options to meet the needs of individual researchers, project teams, and organizations. Contact us to discuss the license that best fits your needs.
DeepMainmast: From Cryo-EM to Atomic Models
DeepMainmast is an automated, AI-powered platform for building protein atomic models from cryo-EM maps. Using an experimental density map and the corresponding protein sequences, it generates structural models for individual proteins and multichain protein complexes.
DeepMainmast applies deep learning to identify amino-acid and backbone-atom features directly within the cryo-EM density. It then traces Cα backbone paths, assigns protein sequences, and assembles compatible fragments into complete atomic models.
The platform can also integrate predicted structures from tools such as AlphaFold. By selecting regions that agree with the experimental density, DeepMainmast helps improve model building in challenging or lower-resolution regions.
CryoREAD: Build DNA and RNA Structures from Cryo-EM Maps
CryoREAD is fully automated, AI-powered software for de novo modeling of DNA and RNA structures from cryo-EM maps. Using an experimental density map and the corresponding nucleic-acid sequences, it generates complete atomic models of DNA and RNA.
CryoREAD applies deep learning to detect nucleic-acid density and identify local structural features, including phosphate, sugar, and base positions. It then traces the nucleic-acid backbone, assigns the target sequences, and assembles compatible fragments into complete atomic models.
This automated workflow is particularly valuable for challenging cryo-EM maps in which DNA or RNA model building is limited by lower or variable local resolution.
DAQ-Score: Validate Protein Structure Models in Cryo-EM Maps
Protein structures built from cryo-EM maps can contain local modeling errors that are difficult to detect using conventional validation methods, including incorrect amino-acid assignments and residue register shifts.
DAQ-Score uses deep learning to evaluate how well each residue in a protein model agrees with its local cryo-EM density. It highlights regions that may be incorrectly modeled, helping researchers identify and correct potential errors and produce more accurate, reliable protein structure models. With a better-validated structural model, researchers can approach downstream applications, including structure-based drug discovery, functional analysis, and protein engineering, with greater confidence.