Our dSPRINT predictor can run on any human Pfam protein domain, and return per-position predictions of interaction-sites for five different ligand types: DNA, RNA, ion, peptide, and small molecule.
If you use data or scripts from this repository, please cite:
Etzion-Fuchs A., Todd D.A. and Singh M., (2021) "dSPRINT: predicting DNA, RNA, ion, peptide and small molecule interaction sites within protein domains", Nucleic Acids Research https://doi.org/10.1093/nar/gkab356
Given a domain input file: 1) all the required external datasets are downloaded, 2) features are calculated for each of the input domain positions, 3) our trained predictors are run and return per-position prediction results.
This repository can be used as a computation pipeline, and uses Snakemake as the underlying engine.
Essentially, given a file input.hmm, with one or multiple domains which follow the syntax of a Pfam-A entry, the following computational graph of rules is run:
Output files are generated in the output folder, with the final result per-position ligand binding score generated
in the file output/binding_scores.csv
ligand_type binding_score domain match_state
0 dna 0.9916359186172485 zf-C2H2 1
1 dna 0.9872528910636902 zf-C2H2 10
2 dna 0.997771143913269 zf-C2H2 11
3 dna 0.997983455657959 zf-C2H2 12
4 dna 0.9957016110420227 zf-C2H2 13
5 dna 0.9956439733505249 zf-C2H2 14
Read the Getting Started guide on how to run dSPRINT.