About Us
Platform
Service
Community
Contact Us
Atombeat
中文
Hermite® — All For Life
New-Generation Drug Design Platform Powered by AI, Physics, and High Performance Computing
About
Hermite® is a new-generation drug computing design platform which integrates artificial intelligence, physical modeling and high-performance computing to provide a one-stop computing solution for preclinical drug research and development. Main functions of Hermite® include structure modeling, drug-target binding mode prediction, virtual screening and lead optimization, etc. Via the form of web access, Hermite® provides an interactive molecular display interface, which supports cross-window intelligent collaboration, various molecular displays and functions, and facilitates users to view, analyze and share proteins and drug molecule structure and simulation data. Meanwhile, Hermite® supports deployments both on premise and on cloud.
Solutions
Structure Modeling
Hermite® provides a variety of solutions in structural modeling and optimization, including Uni-Fold, Uni-AbFold, Uni-EM, etc. Uni-Fold was the first protein structure prediction model in the world that made both inference and training codes open-source. Uni-Fold is open source for business uses, customized computing powers and more accurate on multimer predictions.
Reinforced Dynamics
Uni-RiD uses machine learning to automatically find the set variable that can accurately describe the protein conformation, and realize efficient sampling of potential energy surfaces of protein systems combined with molecular dynamics simulation. Uni-RiD has been successfully used to solve complex sampling problems in scenarios such as protein folding, protein structure optimization, and allosteric pocket prediction, to achieve efficient prediction and optimization of protein dynamics and protein receptor binding conformation.
Docking and Virtual Screening
Hermite® can support high-throughput virtual screening in different scenarios. Uni-Docking, an extremely optimized molecular docking engine based on GPU acceleration, can achieve hundreds of times faster than AutoDock Vina while maintaining consistent precision, and complete high-throughput screening on 12 million ligands with no more than 40 minutes using 100 mounts of V100 cards. Uni-IFD can accurately predict the binding mode of drug molecules and targets by simulating the "induced fit" effect.
Binding Affinity Evaluation
Based on Molecular Mechanics simulations, Hermite® Uni-MM PB/GBSA can help predict the binding energy of ligands in minutes thus can be applied to the screening and sequencing of ten thousand level molecular libraries. Uni-FEP, which combines free energy perturbation theory, molecular dynamics, enhanced sampling algorithms and high-performance computing to efficiently assess the binding affinity of proteins and ligands with chemical precision. Uni-FEP is suitable for R group optimization, core hopping, charge transfer, macrocyclic ring formation and other different optimization scenarios to achieve industrial-scale lead compound optimization.
Key Water Free Energy Calculation
Uni-Aquasite can find the positions of stable and unstable water molecules at the target binding site by calculating free energy, and provide structural basis for drug design.
Molecular Property Prediction
Hermite® Uni-Mol is the first released 3D molecular pre-training model. Uni-Mol directly inputs molecular 3D coordinate information as a model, surpasses SOTA in almost all downstream tasks related to drug molecules and protein pockets, and can directly complete 3D conformation-related tasks such as molecular conformation generation, protein-ligand binding site prediction and druggability prediction, etc.
Biologics Drug Design
Hermite® combines Deep Learning and Molecular Dynamics simulations to provide a complete and efficient solution for antibody discovery, design, modification, and property prediction. Uni-AbFold can accurately predict the 3D structure of antibody, utilize LoopOpt and RiD enhanced sampling functions to further optimize the CDR region.