Use OpenDDE Online
Commercially Available OpenDDE No-Code Web Server
OpenDDE: A New Era for Structural Reasoning and Drug Discovery
Open Drug Discovery Engine (OpenDDE) is an open-source, all-atom biomolecular foundation model that uses co-folding as the entry point to a scalable, AI-driven drug discovery engine. Rather than treating structure prediction as an isolated endpoint, OpenDDE is designed as a shared structural reasoning layer for modeling sequence-structure-function relationships across biomolecular complexes.
By integrating advances in all-atom architecture, atomic latent reasoning, inference optimization, and large-scale data processing, OpenDDE achieves IsoDDE-level co-folding accuracy within a reproducible and openly accessible framework under the Apache-2.0 license. OpenDDE aims to democratize access to frontier biomolecular intelligence, accelerate global collaboration, and lay an open foundation for next-generation drug discovery systems that design, score, and optimize therapeutic candidates for human health.
How OpenDDE Works
OpenDDE is built as a coarse-to-fine molecular structure generation model that unifies structure prediction and conditional design within a single diffusion framework.
Atomic Latent Reasoning: The model incorporates latent reasoning over biomolecular tokens, building residue-token and pair representations to encode hypotheses about molecular geometry. This stage allows the model to refine representations of local geometry, chemical context, and cross-molecular interfaces before all-atom structure generation.
Structural Token Expansion & Refinement: Each residue is expanded into chemically meaningful structural tokens—such as protein backbone, side-chain, nucleic-acid backbone, base, and ligand/atom tokens. A Structural Refiner updates these tokens using pair-conditioned attention and triangular updates, ensuring geometric consistency across indirect token paths before coordinate generation.
Shape-Complementarity Interface Fitting: To guide the model toward physically compatible interfaces, OpenDDE employs geometry-aware objectives. A differentiable shape-complementarity loss compares predicted and native interface geometry by examining surface orientation, spacing, and anti-clash behavior.
Unified Prediction and Design: OpenDDE treats structure prediction and de novo design as instances of the same conditional denoising problem. By adjusting target atom masks, the system can perform full structure prediction or act as a conditional generation engine that designs missing molecular components around fixed structural context.
Performance and State-of-the-Art Benchmarks
OpenDDE delivers exceptional, balanced performance across diverse biomolecular complex prediction tasks, performing on par with leading closed systems and substantially outperforming existing open-source baselines:
Unmatched Antibody-Antigen Accuracy: On the stringent antibody-antigen benchmarks, OpenDDE achieves the highest overall DockQ success rates: 51.0% on PXMeter-AB, 70.0% on FoldBench-AB, and 66.4% on the newly curated 2026ARK-AB benchmark, vastly outperforming models like ESMFold2, Protenix-v1, Boltz-1, and Chai-1.
Robust Generalization to New Biology: Evaluated on low-homology complexes released in 2026, OpenDDE demonstrates strong generalization capability, accurately recovering experimental binding geometry and epitope-facing poses across membrane proteins and novel biological assemblies.
Predictable Test-Time Scaling: OpenDDE benefits predictably from increased inference-time sampling budgets. When generating multiple stochastic samples with different random seeds, the oracle success rate rises from roughly 66% with a single seed to above 90% with hundreds of seeds.
Unprecedented Compute & Parameter Scale: OpenDDE enters the language-model style scaling regime with 655 million parameters trained across 414,000 GPU-hours. It reaches the highest protein monomer accuracy with an LDDT score of 0.890 and a top protein-protein score of 0.769 on FoldBench.
What is Tamarind Bio?
Tamarind Bio is a pioneering no-code bioinformatics platform built to democratize access to powerful computational tools for life scientists and researchers. Recognizing that many cutting-edge machine learning models are often difficult to deploy and use, Tamarind provides an intuitive, web-based environment that completely abstracts away the complexities of high-performance computing, software dependencies, and command-line interfaces.
The platform is designed to provide easy access to biologists, chemists, and other researchers who may not have a background in programming or cloud infrastructure but want to run experimental models with their data. Key features include a user-friendly graphical interface for setting up and launching experiments, a robust API for integration into existing research pipelines, and an automated system for managing and scaling computational resources. By handling the technical heavy lifting, Tamarind empowers researchers to concentrate on their scientific questions and accelerate the pace of discovery. The Tamarind team holds information/data security as a top priority, as detailed in our Trust Center & Terms of Service, ensuring your data is safe on the platform.
How to Use OpenDDE on Tamarind Bio
Leverage the all-atom modeling power of OpenDDE through Tamarind Bio’s streamlined, no-code web environment:
Access the Platform: Log in to your account on the app.tamarind.bio website.
Select OpenDDE: From the available suite of cutting-edge computational tools, choose the OpenDDE tool.
Configure Your Input: Provide your sequence, atom, template, or MSA features. You can input your query for standard all-atom structure prediction or define known-target masks if you are performing structure-conditioned de novo design.
Set Your Sampling Budget: Define your inference-time parameters, choosing the number of random seeds or stochastic samples to generate to leverage OpenDDE's favorable test-time scaling.
Run the Denoising Engine: Launch the experiment. OpenDDE scales seamlessly using context-parallel layouts (Fold-CP) to manage large biomolecular systems efficiently without memory bottlenecks.
Evaluate Outputs: Download the fully denoised all-atom coordinates and analyze the localized structural configurations, interface geometries, and confidence/distance head prediction outputs generated by the model.