Structure prediction

OpenDDE

Structure prediction for proteins, DNA/RNA, ligands, and ions. Supports co-folding.

OpenDDE is an all-atom biomolecular foundation model that turns co-folding into a scalable engine for structure prediction, design, and optimization in drug discovery. It models proteins, nucleic acids, and small molecules in one all-atom system.

Documentation ↗ Input parameters ↗ Examples ↗ Home page ↗ Paper ↗ Source code ↗ License ↗

No jobs before this; it will run immediately.

Not logged in. Register an account to keep track of your jobs and results, and add presets and bookmarks. Email results to this address when complete:

Ready

Example preset

Choose an example to fill in its settings and input structures.

Input mode

Input

Job name Names the single OpenDDE job generated from these parameters.

Molecules

Add one box per unique OpenDDE entity. Set count and an optional matching list of comma-separated IDs, plus the sequence/ligand/ion, type-specific modifications, and optional MSA or template paths.

Covalent bonds (JSON, optional) Native OpenDDE covalent_bonds array. Entity references are 1-based sequence indexes; copy references are 1-based within count.

OpenDDE inference jobs (JSON) Complete native OpenDDE input: a non-empty top-level list of named jobs using proteinChain, dnaSequence, rnaSequence, ligand, or ion entities. In this form: Bio Web writes this text to input.json. Any MSA, template, or ligand-file paths inside it must be absolute paths available to the runner.

Inference jobs file (JSON) The JSON file defining this run's jobs, in OpenDDE's own input format. In this form: Choose a file. Its contents are read here and sent as the input document.

Inference

Override model seeds (optional) Comma-separated integers. When set, --seeds overrides every job's modelSeeds; otherwise OpenDDE uses modelSeeds from the JSON or samples a random seed.

Samples Number of structure samples per seed. The OpenDDE CLI default is 5.

Diffusion steps Number of diffusion steps. The OpenDDE CLI default is 200.

Pairformer cycles Number of Pairformer recycling cycles. The OpenDDE CLI default is 10.

Runtime

Inference precision OpenDDE defaults to FP32; BF16 can reduce memory use on supported devices.

Device Auto selects CUDA when available and otherwise CPU.

Optional features

Use or generate protein MSAs Use protein MSA features. A protein without paired or unpaired MSA paths has its sequence sent to the public ColabFold MMseqs2 server (https://api.colabfold.com; set MMSEQS_SERVICE_HOST_URL on the runner to use your own). If that search fails, OpenDDE continues with a query-only MSA and the result lists a warning. Turn off for private sequences or offline runs, at a significant cost in accuracy.

Use protein templates Use template features from each protein's templatesPath. A protein without one gets a local template search instead, which needs protein MSAs, HMMER (hmmsearch and hmmbuild) on the runner's PATH, and downloads the ~220 MB PDB seqres database on first use.

Use RNA MSAs Use RNA MSA features from each RNA's unpaired MSA path. An RNA without one gets a local nhmmer search instead, which needs HMMER on the runner's PATH and downloads roughly 90 GB of RNA databases on first use. Requires protein MSAs to be on.

Use training-free guidance Apply OpenDDE's geometry-based Training-Free Guidance to protein-ligand sampling.

Output

Write atom-level confidence data Write OpenDDE's full per-atom confidence JSON in addition to summary confidence.

Runtime

Deterministic PyTorch algorithms Ask OpenDDE to use deterministic PyTorch algorithms for reproducible inference.

Ready