spaed 1.0.3

A module for the segmentation of phage endolysin domains based on the PAE matrix from AlphaFold.

Segmentation of PhAge Endolysin Domains

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.

Installation & usage

Check out www.spaed.ca to launch SPAED quickly!

First create a virtual environment, then:

From pypi:

pip install numpy, pandas, scipy, 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

Advanced usage

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/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 into a new fasta file named "spaed_predicted_domains.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/disordered regions. Residues with PAE score < PAE_SCORE_CUTOFF are considered close together. (default = 4).
• MIN_DISORDERED_SIZE - Minimum size a 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 disordered 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 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).

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 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.

Outputs

A csv file containing the proteinID, length, number of predicted domains, domain delineations, linker delineations, 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

Citation

Boulay, A. et al. SPAED: Harnessing AlphaFold Output for Accurate Segmentation of Phage Endolysin Domains. 2025.04.25.650745 Preprint at https://doi.org/10.1101/2025.04.25.650745 (2025).