Use RoseTTAFold2

The amino acid sequence for the protein structure you wish to predict. To predict complexes / multimers use ":" to specify inter-protein chain-breaks. For example PISK:KDIP for a heterodimer.

Select the symmetrical design you want to use to predict your structure. Leave this as Unspecified if you don't require any fancy symmetries. Dihedral will correspond to 2 * Order * Sequence.

Defines the number of copies in the symmetries. Think of it was the desired number of subunits in the predicted structure. This option will be ignored for Tetrahedral, Octahedral, and Icosahedral.

Select the method MSAs are concatenated. Diagonal concatenation preserves relative geometry and domain alignment of mulit-domain proteins. Repeat concatenation forms an extended MSA, use with caution. Default is what was used in RoseTTAFold2 preprint version results. Diag mode was found to give the best results so it is what we recommend.

Select mmseq2 for the ColabFold method or single_sequence for de-novo design where a homologous sequence will not be found. Warning: We query a 3rd party service to perform sequence searches, if you do not wish to share your sequence set this to "single_sequence".

The unpaired mode generates separates MSA for each protein, unpaired_paired attempts to pair sequences from the same genome, and paired only uses successfully paired sequences.

Select the number of recycles you want to use. Increasing recycling steps tends to produce more accurate predictions but will also greatly increases prediction time. For smaller proteins and monomers we recommend 5 recycles, for bigger proteins (greater than 1,500 AA) we recommend 15 recycles. If you need to increase this value beyond 15 contact customer support.

Check this option to randomly mask 15% of positions in the input MSA.

Check this option to enable dropout layers during inference.

If max_msa * 8 is lower than number of sequences in the input MSA. The number of sequences are subsampled to the max number. Different subsamplings may also results in stochasticity.

Provide a random seed to allow reproducible stochasticity.