OmegaFold: High-resolution de novo structure prediction from primary sequence

OmegaFold is a protein structure prediction algorithm that uses a protein language model to predict structure. The algorithm works without multiple sequence alignments, which means OmegaFold runs faster than AlphaFold and also works on divergent sequences (sequences without many homologs).

Why should I use OmegaFold?

Unlike AlphaFold (and ColabFold), OmegaFold does not rely on multiple sequence alignments during prediction. This makes OmegaFold potentially better suited for proteins that have low sequence coverage. Note that multiple sequence alignments may help OmegaFold, but at this moment, COSMIC2 does not support this type of modification.

OmegaFold can predict up to 4096 amino acids on a single GPU. If you run into memory limitations, use the subbatch size (see below)

Running OmegaFold on COSMIC²

Accessing COSMIC2: https://cosmic2.sdsc.edu

Input: a FASTA protein sequence file containing your sequence of interest.

  • Upload data via browser upload (not Globus!)
  • The first line of the FASTA file will be the output file name (“> My favorite protein” becomes “My_favorite_protein.pdb”)
    • If this name has special characters (e.g. “=”, ” | “, etc) this will interfere with downloading results from COSMIC2, so adjust accordingly.

Options:

  • Number of cycles: [Default = 10]
    • The number of cycles for optimization
  • Subbatch size: [Default = -1, whole sequence]
    • Length of sequence to use during prediction. By default, OmegaFold will use the full sequence, but larger sequences may run out of memory. If the job crashes due to memory limitations, indicate the maximum sequence length to use during prediction.

Predicting protein-protein complexes

OmegaFold does not work on multiple input sequences. Instead, you can insert a series of glycine-serine linkers between the proteins of interest

Checking job status

After you submit your job, you can check your job status by clicking on the hyperlink ‘List’ which is next to the label ‘Intermediate Results.’ This will open a new window that lists the current job directory to show you files as they are generated. Watch a video here.

OmegaFold outputs on COSMIC²

OmegaFold produces a single structure, which is available for download on the output page.

Citation:

High-resolution de novo structure prediction from primary sequence
Ruidong WuFan DingRui WangRui ShenXiwen ZhangShitong LuoChenpeng SuZuofan WuQi XieBonnie BergerJianzhu MaJian Peng