Use AlphaFold2

Amino acid sequences of the structures you want to predict in FASTA format. Note that each sequence must have a unique alphanumeric name. To predict complexes / multimers use ":" to specify inter-protein chainbreaks. For example PI...SK:PI...SK for a homodimer. Additionally, gaps within sequences denoted with either "-" or "." are automatically removed. Trailing stop codons are also automatically removed, but a premature stop codon is not allowed. Maximum 2 sequences for free plan, maximum of 10 for paid plans.

Optionally provide a custom template structure (pdb or mmCIF formats) for AlphaFold2 to use during prediction. Note that custom templates can only be used if you are predicting the structure of a single protein.

Optionally provide a custom MSA (fasta or a3m format) for AlphaFold2 to use during the prediction. Note that custom MSAs can only be used if you are predicting the structure of a single protein. Make sure the first sequence in the MSA is the one you want to predict the structure of. Sequences will be truncated to the length of the first sequence (sequence that gets folded).

Check this option to utilize amber force fields to relax the predicted structure. This doesn't improve the prediction accuracy but does help remove otherwise distracting stereochemical violations.

Select the model you want to use to predict your structure. For monomers we recommend alphafold2_ptm, for complexes we recommend alphafold2_multimer_v3. If you leave this as auto alphafold2_ptm will be used for monomers and alphafold2_multimer_v3 will be used for complexes.

A template is an optional input structures that should be similar, but different to the structure you are trying to predict. AlphaFold2 uses templates as a sort of "guide" to help it improve its prediction. Using the "none" option will result in no template being used, the "pdb70" option results in a similar structure from the pdb70 being used. Note if you upload a custom template then your selection here will be ignored. Warning: We query a 3rd party service to perform template searches, if you do not wish to share your sequence do not enable this option.

Choose the MSA database you want to use. Selecting "single_sequence" will result in the MSA containing only your target sequence. Note if you upload a custom MSA then your selection here will be ignored. In most cases mmseqs2_uniref_env tends to produce the best results. 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".

This option applies to the input MSA. "unpaired_paired" = pair sequences from same species + unpaired MSA, "unpaired" = separate MSA for each chain, "paired" - only use paired sequences. Pairing the MSA can help AlphaFold2 create better predictions for complexes.

Increased recycling steps tend to produce more accurate predictions but will also greatly increases prediction time. For smaller proteins and monomers we recommend 5 recycles, for bigger proteins we recommend 10 recycles. Maximum of 8 recycles for free users. Upgrade your plan to increase recycling values.

Ends the protein folding process if the improvement between recycling steps doesn't surpass the tolerance value. The tolerance is defined as the RMSD (difference in distance matrices, angstrom units) between recycles. If it drops below the specified value, the recycling will terminate. We recommend a tolerance of 0.3 for smaller proteins and 0.5 for larger proteins and complexes.

The trunk of the network is run multiple times with different random choices for the MSA cluster centers (1=default, 8=casp14 setting). Increasing ensembling can occasionally improves prediction quality but will GREATLY increase prediction time. Maximum of 1 ensembles for free users. Upgrade your plan to increase ensembles values. If you need to increase this value beyond 3 contact customer support. CURRENTLY DISABLED DUE TO HIGH DEMAND. IF YOU NEED THIS VALUE INCREASED PLEASE CONTACT SUPPORT.

Check this option to enable "Training Mode" which activates drop out layers within the model. This can be used to force the model to create more diverse predictions.