OpenDDE co-folding
All-atom co-folding for any mix of protein, DNA, RNA, and ligand in one spec. AlphaFold3-class multi-modal structure prediction, the multi-molecule complement to the protein-only Boltz-2 tool.
What it is for
Pick OpenDDE to co-fold a mixed complex — protein with DNA, RNA, or a bound ligand in one prediction. For a plain protein-protein or protein-peptide cofold, Boltz-2 is faster and cheaper.
OpenDDE is an AlphaFold3-class, all-atom co-folding foundation model (Aureka AI Research, Apache-2.0). It predicts the joint structure of an arbitrary mix of biomolecular entities — protein, DNA, RNA, and small molecules (ligands) — from a single specification.
When it fits:
- You need a complex with more than just protein: protein plus DNA / RNA, or a bound small molecule.
- You are modelling an antibody or nanobody with its antigen (use the ABAG checkpoint).
- You want an AlphaFold3-style multi-modal prediction without standing up the pipeline yourself.
Inputs
You will need:
- Sequences for each polymer chain (protein / DNA / RNA).
- For ligands: a CCD code (e.g.
CCD_ATP), a bare SMILES string, or a bundledFILE_*.sdfreference.
Each run uses a preset that sets the scale and scope:
- General co-folding
- The general OpenDDE checkpoint. Co-fold any mix of protein, DNA, RNA, and ligand entities. Choose this unless your target is specifically an antibody-antigen pair.
- Antibody-antigen (ABAG)
- The antibody-antigen specialised checkpoint. Tuned for predicting antibody or nanobody complexes with their antigen.
Parameters you set on the form:
- Checkpoint
- General for any entity mix, or ABAG for antibody-antigen complexes.
- Entities (guided)
- One textarea per entity type. Proteins / DNA / RNA are entered as FASTA (
>idheaders optional); ligands one per line. This adapter assembles the OpenDDE JSON for you. - Entities (JSON)
- Or paste an exact OpenDDE spec. It is validated against the real schema and re-checked against the same size limits — the JSON mode is not a way around the ceilings.
- Samples / steps / recycles
- Sampler settings. Samples is how many structures per seed; steps and recycles trade compute for quality.
- Seeds
- How many seeds to run from the starting seed. Total predictions returned is seeds × samples.
Typical runtime:
- general
- ~2 to 8 min
- abag
- ~2 to 8 min
How to read the results
A ranked set of predicted complexes (mmCIF/PDB) with the model's own confidence ranking score per prediction. Download each structure or view it in the browser.
Where a tool reports them, the scores mean:
- ipTM
- Predicted confidence in the binder to target interface. Higher is better. Aim above roughly 0.7 on a tractable target.
- pLDDT
- Per-residue confidence in the predicted fold. Higher means the model is more sure of that part of the structure.
- i_pAE and pAE
- Predicted alignment error, at the interface (i_pAE) or across the whole structure (pAE). Lower is better.
References
Aureka AI Research, OpenDDE-Preview, arXiv 2026