ProSSpeC - Protease Substrate Specificity Calculator

NIa Protease specificity modeling where Direct Coupling Analysis (DCA) is used to learn couplings and local fields, which are then used to calculate the Hamiltonian specificity between proteases and substrates. An associated web app to run these calculations is available on https://www.coevolutionary.org/prosspec/.

dca.m is used to calculate direct information (DI) values
DCAparameters.m is used to calculate the local fields (h_i) and couplings (e_ij)
FastaHamiltonian.m is used to calculate the raw Hamiltonian values.
FastaHamiltonianTARGET.m is used to calculate the raw Hamiltonian values for masked sequences.
specificity.m is used to construct the Hamiltonian specificity.

To run:

Install Matlab (test on 2024a) and HMMER (v3.4). (Takes less than 10 minutes)
Download DCA.zip files from 10.5281/zenodo.15039890
Run specificy.m using the unzipped parameters from DCA.zip on a protein sequence fasta.
The protein sequence fasta must be first aligned to the protease and substrates. Use HMMER and the files from HMM.zip in Zenodo to perform the alignments. The concatenate the protease and substrate (in that order).

Note: We recommend using the webserver as it performs the alignments and calculations for you.

Demo:

1. Download files from Zenodo
2. Unzip DCA.zip and move parameter files into \demo

unzip DCA.zip
mv *.mat ProSSpec/demo

3. Run via matlab from command line. This demo is for Ryegrass mosaic virus.

matlab -nodisplay -nosplash -r "cd('matlab scripts'); disp(specificity('../demo/demo.fasta', '../demo/concatenated_sequences_DCAparams.mat', '../demo/scrambled_sequences_DCAparams.mat')); exit"

4. Compare to the expected output ( -493.9972 ).

Reproducability:

1. Model can be retrained using the fastas in Zenodo DCA.zip folder and DCAparameters.m.
2. Sequences can be retrieved via Uniprot and/or imported from Plasmids.zip and then aligned using hmmalign from HMMER and the provided hmm profiles in Zenodo HMM.zip.
3. Once sequences are preprocessed, concatenate them with the protease sequence first, then substrate. Keep in fasta format.
4. Rerun inference on these sequences and compare to the outputs in Zenodo Hamiltonians.xlsx.

To cite:

Identification and engineering of highly active Potyviral proteases using co-evolutionary models. Medel B. Lim Suan Jr., Cheyenne Ziegler, Zain Syed, Ajay Tunikipati, Rodrigo Raposo, Jaimahesh Nagineni, Jaideep Kaur, Faruck Morcos, and P. C. Dave P. Dingal