RoseTTAFold3

Predict the structure of a complex of proteins, DNA, RNA and ligands with RoseTTAFold3. Each sample comes with a ranking score, pTM, ipTM, per-residue pLDDT and predicted aligned error, so it can serve as an independent check next to Boltz-2 and ESMFold2, for example on designed sequences. Chains are folded from their sequences alone unless you give an alignment, which suits de novo proteins.
5 samples
Estimated minimum cost: $—

Examples

Complex chains

Build one complex per job from protein, DNA and RNA chains and ligands. Give identical copies of a chain several IDs (for example A, B). Write a modified residue as its CCD code in parentheses, such as (MSE). The polymer chains may hold at most 1000 residues in total, copies counted.

Protein chain
or drop a file here
Optional. Without one the chain is folded from its sequence alone.

Sampling and options

Structures drawn from one pass, 1 to 10; ranked by confidence.

Trunk recycles, 1 to 10.

20 to 400; samples × steps is at most 1600.

Set it to reproduce a run; leave it empty for a random seed.

0 to 1. Stops after the first recycle if the mean pLDDT is lower and writes only the score, no structure. Leave empty to always get a structure.

Searches UniProt Swiss-Prot with MMseqs2 for each protein chain that has no uploaded alignment. The alignments are shallow, so natural proteins may fold better with an uploaded one. Leave it off for de novo proteins, which have no homologs.

One bond per line, two atoms written CHAIN/RESNAME/RESID/ATOM, for example A/ASN/133/ND2 F/NAG/1/C1. At most 50.

IDs of protein chains to close into a ring (cyclic peptides), separated by commas.