PPI-Miner is a protein-protein interaction (PPI) search pipeline for proteins containing specific substructure motif (3D) or sequence motif (2D). PPI-Miner can be used to 1) mining potential PPI partners for specific receptors, such as E3, kinase, phosphatase. 2) searching protein scaffolds fusing specific motifs such as antigenic epitopes.
A number of proteins can bind to the same receptor at the same site with similar binding patterns. Most of the time, these proteins belong to the same protein family and have a high degree of sequence similarity. However, these proteins can sometimes be scattered among a diverse family of proteins, maintaining sequence conservation for a short conserved sequence (shown in Figure 1A) or even no apparent sequence conservation (shown in Figure 1B). For one thing, the short conserved sequence between the disordered protein and the receptor is called sequence motif, such as the TRAF6 and its substrates. For another thing, some small sub-structures, called structural motifs, are no apparent sequence conservation and scattered among different protein families, such as CRBN and its substrate proteins with different 3D overall structures. The development of PPI-Miner was inspired by these PPIs that are difficult to search by sequence alignment methods(e.g. Blast).
Figure 1. Kinds of PPI protein motifs.
PPI-Miner needs to be installed on the Linux system. Although the webserver is available at here, we highly recommend that users install the PPI-Miner system locally to use the more complete capabilities.
a. (Required) Intsall the GNU parallel, Julia and Python3.
Parallel:
sudo apt install parallel
Python3 (by miniconda):
wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
sh Miniconda3-latest-Linux-x86_64.sh
Julia:
wget https://julialang-s3.julialang.org/bin/linux/x64/1.5/julia-1.5.3-linux-x86_64.tar.gz
tar zxvf julia-1.5.3-linux-x86_64.tar.gz
cd julia-1.5.3/bin
echo "export PATH=${PWD}:\$PATH" >> ~/.bashrc
source ~/.bashrc
julia
] # in Julia REPL
add BioStructures # in Julia REPL
add PDBTools # in Julia REPL
exit()
b. (Required) Intsall MASTER packages and add to PATH environment variables.
c. (Optional) Intsall the Rosetta packages and set up environment variables, which used to calculate the PPI binding energy. We recommend to install the Rosetta.
echo "export rosetta_app=<path to rosetta apps>" >> ~/.bashrc
echo "export rosetta_db=<path to rosetta database>" >> ~/.bashrc
echo "export rosetta_version=<rosetta versions, mpi or static>" >> ~/.bashrc
d. (Optional) Intsall the EMBOSS packages and add to PATH environment variables, which used to search the sequences by PFM (In most cases, we recommond the sequence parttern rather than PFM).
git clone https://github.com/Wang-Lin-boop/PPI-Miner/
cd PPI-Miner/scripts/
echo " # Scripts of PPI-Miner ,added by PPI-Miner, refer to https://github.com/Wang-Lin-boop/PPI-Miner/
export PATH=${PWD}:\$PATH" >> ~/.bashrc
source ~/.bashrc
1. Download the library of prepared human protein structures and disorder sequeneces.
First, download the 3DPPI-Miner structure library, which may take a bit longer.
wget https://bailab.siais.shanghaitech.edu.cn/service/3DPPI-Miner-pds-Library.zip
unzip 3DPPI-Miner-pds-Library.zip
for pds in `ls 3DMotif-Dock-pds-Library`;do
echo "${PWD}/3DMotif-Dock-pds-Library/${pds}" >> 3DMotif-Dock-pds-Library.list
done
echo " ## Database of 3DPPI-Miner, added by PPI-Miner, refer to https://github.com/Wang-Lin-boop/PPI-Miner/
export HumanProteinPDS=${PWD}/3DMotif-Dock-pds-Library.listt" >> ~/.bashrc
Next, switch to the PPI-Miner/Database and execute the following command.
echo " ## Database of 2DPPI-Miner, added by PPI-Miner, refer to https://github.com/Wang-Lin-boop/PPI-Miner/
export HumanDisorderSequence50=${PWD}/Human-Disorder50-protein.fasta
export HumanDisorderSequence70=${PWD}/Human-Disorder70-protein.fasta" >> ~/.bashrc
Finally, three environment variables will be added to your ~/.bashrc file.
2. Use the 3DPPI-Miner to generate your own structure libraries.
First, you need to build your own library of single chain protein structures, which may be obtained from PDB (please refer to GetPDB), AlphaFold DB (recommend to pre-precessed by Post-AlphaFold) or some conformational ensemble generated by molecular simulation. Next, run the following command to generate the processed structure library.
3DMotif-Dock -i <INPUT_PDB_LIB> -n 40 -1
You can adjust the '-c' option to change the default chainname "B" of database structures, which aim to aviod chainname conflict. For multiple chains PDB library, you can use the '-k' option during next step to change receptor chainname to "Z". The program generates a file named INPUT_PDB_LIB-Index, which will be used in the next calculations as the database index file.
3. Use the scripts to generate your own disorder sequences libraries.
Please refer to AlphaFold Processing.
2DMotif-Dock is the running script of 2DPPI-Miner, Run 2DMotif-Dock -h to show the help information of 2DPPI-Miner.
1. Search for disordered human protein sequence library using sequence pattern.
2DMotif-Dock -i ${HumanDisorderSequence70} -M "..[PG]EE[TS]." -r <receptor.pdb> -c "B"
2. Search for disordered human protein sequence library using MSAprofit (need to EMBOSS).
2DMotif-Dock -i ${HumanDisorderSequence70} -m <msa.fasta> -p 65 -r <receptor.pdb> -c "B"
Expected Output of 3DPPI-Miner:
1. *-ddG.sc: The final socre file, which contains sequences and corresponding ddG score.
2. ddG_running: The mutated complex structures.
3DMotif-Dock is the running script of 3DPPI-Miner, Run 3DMotif-Dock -h to show the help information of 3DPPI-Miner. Take care! don't use the same chain name to your database (default "B"). If it was, you can use the '-k' option to change receptor chainname to "Z" to aviod chainname conflict.
1. Search for flexible structural motif.
3DMotif-Dock -m Motif_1.pdb -r Receptor.pdb -l ${HumanProteinPDS} -n 40 -d 1.0 -2
2. Search for stable structural motif.
3DMotif-Dock -m Motif_1.pdb -r Receptor.pdb -l ${HumanProteinPDS} -n 40 -d 0.6 -2
3. Perform motif searching and optimizing the searched structure (Required rosetta).
3DMotif-Dock -m Motif_1.pdb -r Receptor.pdb -l ${HumanProteinPDS} -n 40 -d 0.6
Expected Output of 3DPPI-Miner:
1. INPUT-Motif-FINAL_interface_score_OUT.sc: The final socre file, the dG_cross_Total is the most recommended criteria.
2. INPUT-Motif-Jd2-OUT: This dir contains the final complexes of your receptor structure with the structure which is in your INPUT lib as well as matched to your motif structure.
3. INPUT-Motif-Relax: The refined complexes for INPUT-Motif-Jd2-OUT.
4. INPUT-Motif-Jd2-?(Target_Chainname).fasta: The sequences of matched motif structures.
L. Wang, F.-l. Li, X.-y. Ma, Y. Cang, F. Bai, PPI-Miner: A Structure and Sequence Motif Co-Driven Protein–Protein Interaction Mining and Modeling Computational Method. Journal of Chemical Information and Modeling, (2022).
We sincerely appreciate the Chinese Rosetta Community, Yuan Liu and Wei-kun Wu, who provide technical communication for this study. We also sincerely thank Joe Greener, the developer of BioStructures, for his help to our Julia code. We are also grateful for the support from HPC Platform of ShanghaiTech University.