Method to infer PTM (currently only phosphorylation) hotspots in conserved protein alignments based on (Strumillo et al. bioRxiv).
ptm_hotspot requires Python (v3) as well as the following packages:
- numpy
- pandas
- scipy
- statsmodels
If you want to predict all hotspots in all protein domains just run:
python3 ptm_hotspots.py -o predicted_hotspots.csv
To obtain particular domain predictions:
python3 ptm_hotspots.py -o kinase_domain_hotspots.csv -d PF00069
To obtain hotspot residue predictions instead of hotspot ranges
python3 ptm_hotspots.py -o hotspot_residues.csv --printSites
You can obtain more help and options by typing:
python3 ptm_hotspots.py -h
usage: ptm_hotspots.py [-h] [--dir [PATH]] [--ptmfile [PATH]] [-d [PFXXXXX]]
[--iter [INTEGER]] [--threshold [FLOAT]]
[--foreground [FLOAT]] -o PATH [--printSites]
Estimate PTM hotspots in sequence alignments
optional arguments:
-h, --help show this help message and exit
--dir [PATH] fasta alignments dir (default: db/alignments)
--ptmfile [PATH] file containing PTMs (default: db/all_phosps)
-d [PFXXXXX], --domain [PFXXXXX]
query single domain (i.e. PF00069)
--iter [INTEGER] number of permutations (default: 100)
--threshold [FLOAT] Corrected p-value threshold (default: 0.01)
--foreground [FLOAT] effect-size foreground cutoff (default: 2)
-o PATH, --out PATH output csv file
--printSites print all residue predictions
Note: Since the Bonferroni correction depends on the total number of predictions, small disimilarities might emerge in the same domain hotspots depending on whether you run only a domain or the full set of domains. Similarly, the stochastic nature of the permutation analysis might make the results vary between runs.
By default ptm_hotspot uses a database containing precalculated domain alignments (as described in Strumillo et al.) as well as a collection of phosphorylated residues derived from public high-throughput mass spectrometry experiments. In order to update the database please consider the next points:
Every alignment file should be in FASTA format and the header should contain the start and the end of the domains in the alignment coordinates separated by ";". For example:
>EDP05298 pep:known supercontig:v3.1:DS4 ;51;337
For full protein predictions just include the first and last positions in the multiple sequence alignment.
The ptm database should be included as a csv file containing id, amino acid and position of the phosphosite within the protein.
- Strumillo, M. J., Oplova, M., Viéitez, C., Ochoa, D., Shahraz, M., Busby, B. P., et al. Sopko, M., Studer, R. A., Perrimon, N., Panse, V. G., Beltrao, P. (2018). Conserved phosphorylation hotspots in eukaryotic protein domain families. bioRxiv. https://doi.org/10.1101/391185