SPAED is a tool to identify domains in phage endolysins. It takes as input the PAE file(s) obtained from AlphaFold and outputs a csv file with delineations.
Additional scripts are provided to visualize predicted domains with PyMOL and to obtain their amino acid sequences.
Check out www.spaed.ca to launch SPAED quickly!
First create a virtual environment, then:
From pypi:
pip install spaed ### note the spelling of spaed
ex. spaed pae_path --output_file spaed_predictions.csv
From source:
git clone https://github.com/Rousseau-Team/spaed.git
pip install numpy pandas scipy
ex. python spaed/src/spaed/spaed.py pae_path
Optional dependency for structure visualisation: pymol (conda install -c conda-forge -c schrodinger pymol-bundle). Python>3.10 is required, 3.12.9 worked for me.
ex. (install from pip). pymol_vis pred_path pdb_path --output_folder pymol_output --output_type {pse|png|both}
ex. (install from source). python spaed/src/spaed/pymol_vis.py pred_path pdb_path --output_folder pymol_output --output_type {pse|png|both}
Positional arguments:
- pae_path - Folder of or singular PAE file in json format as outputted by Alphafold2/3 or Colabfold.
Optional arguments:
- output_file - File to save table of segmented domains in csv format. (default spaed_predictions.csv)
- fasta_path - Path to fasta file or folder containing fasta files. If specified, spaed will save the sequences corresponding to predicted domains,linkers and terminal disordered regions into new fasta files named "spaed_predicted_{seq_type}.faa" in the same output folder as output_file. Ensure fasta names or headers correspond to entries in pae files.
- RATIO_NUM_CLUSTERS - Maximum number of clusters initially generated by hierarchical clustering corresponds to len(protein) // RATIO_NUM_CLUSTERS. (Default 10). For a protein 400 residues long, 40 clusters will be generated.
- MIN_DOMAIN_SIZE - Minimum size a domain can have. (default 30).
- PAE_SCORE_CUTOFF - Cutoff on the PAE score used to make adjustments to predicted domains/linkers/terminal disordered regions. Residues with PAE score < PAE_SCORE_CUTOFF are considered close together. (default = 4).
- MIN_DISORDERED_SIZE - Minimum size a terminal disordered region can be to be considered a separate entity from the domain it is next to (default 20).
- FREQ_DISORDERED - For a given residue in the PAE matrix, frequency of residues that can align to it with a low PAE score and still be considered "not part of a domain". Values <MIN_DOMAIN_SIZE are logical, but as it increases, the more leniant the algorithm becomes to non-domain regions (more will be predicted). (default 6).
- PROP_DISORDERED - Proportion of residues in a given region that must meet FREQ_DISORDERED criteria to be considered a terminal disordered region. The greater the value, the stricter the criteria to predict the region as disordered. (default 80%).
- FREQ_LINKER - For a given residue in the PAE matrix, frequency of residues that can align to it with a low PAE score and still be considered as part of the linker. Values < MIN_DOMAIN_SIZE are logical as they are less than the expected size of the nearest domain. Increasing leads to a more leniant assignment of residues as part of the linker. (default 20).
- version - Display installed SPAED version number.
If you are interested in looking at the disordered regions in N- or C-terminal, consider increasing FREQ_DISORDERED ([4-30]), decreasing MIN_DISORDERED_SIZE ([10-30]) or decreasing PROP_DISORDERED ([50-95]). This will result in more (and longer) terminal disordered regions being detected, but also many false positives. I would not change them all at the same time as this will probably increase the sensitivity too much.
If you are interested in linkers or have a protein that is less well folded, consider modifying the FREQ_LINKER parameter ([4-30]). This value is used to adjust the boundaries of the linkers and as such, a higher value will result in longer linkers. However, linkers that were missed will still not be detected.
A csv file containing the proteinID, protein length, number of predicted domains, domain delineations, linker delineations, terminal disordered region delineations. Delineations for each domain are separated by a ";".
Ex.
| length | # domains | domains | linkers | disordered | |
|---|---|---|---|---|---|
| prot 1 | 251 | 2 | 1-120;130-251 | 121-129 | |
| prot 2 | 386 | 2 | 86-203;217-386 | 204-216 | 1-85 |
Alexandre Boulay, Emma Cremelie, Clovis Galiez, Yves Briers, Elsa Rousseau, Roberto Vázquez, SPAED: harnessing AlphaFold output for accurate segmentation of phage endolysin domains, Bioinformatics, Volume 41, Issue 10, October 2025, btaf531, https://doi.org/10.1093/bioinformatics/btaf531.