minnie - a structural ensemble analysis package to deduce the fingerprints of binding
developed by Deniz Dogan and Ezgi Karaca
The molecular dynamics (MD) simulations produce extremely valuable information on the functionally important interactions. A prominent way to define these interactions relies on the characterization of inter-molecular contacts that are persistently similar or different across different simulation trajectories. Expanding on this notion, we have developed the python package, minnie (Molecular INteractioN fIngErprints), which:
(i) calculates inter-/intra-molecular hydrogen bonds, ionic, hydrophobic and ring stacking interactions by using interfacea python package; (ii) filters out the non-essential interactions via a user-defined observation frequency filter; (iii) compares the common and distinct interactions across given simulation sets; (iv) outputs the common and distinct interaction profiles in table format and compares their distribution with box-and-whisker plots.
minnie compares the simulation sets of any complex type (i.e. protein-protein/protein-DNA/protein-ligand complex). As long as the residue numbering is comparable, minnie finds similar and distinct interactions between two simulations of the same system, between two highly homologous systems, as well as between wild-type and mutant states of the complex under study.
git clone https://github.com/CSB-KaracaLab/minnie.git
or, if you don't use git:
wget https://github.com/CSB-KaracaLab/minnie/archive/master.zip
minnie functions under the umbrella of anaconda3 (with Python 3.6). So, before everything, anaconda3 should be installed.
cd minnie
conda env create -f environment.yml
conda activate minnie
cd interfacea
python setup.py build
python setup.py install
cd ..
bash execute.sh
setup.sh script installs the required dependencies together with the first version of interfacea. setup.sh finally calls execute.sh script, which sets proper aliases and updates your .bashrc file.
When the environment is set
source ~/.bashrc
1) Your trajectory should be saved in an ensemble format while different conformers are separated by ENDMDL. Then splitpdbs option of minnie is used to split your trajectories into single frames.
❗ Please be sure that your trajectory also includes chainIDs. This is essential to dissect complex-specific interactions at the next steps.
❗ If your input structures do not come from MD simulation, make sure that they don’t have any missing heavy atoms/hydrogens.
The resulting frames are saved under projectID1/01_frames & projectID2/01_frames.
minnie splitpdbs -i projectID1 projectID2 --pdbs ensemble1.pdb ensemble2.pdb
In our toy model (as provided under example_run), the project id folders (-i, i.e. complex names) are defined as sox4 & sox18, where the ensembles are fed as sox4.pdb & sox18.pdb.
minnie splitpdbs -i sox4 sox18 --pdbs sox4.pdb sox18.pdb
2) When the individual frames are generated in your project folders, their inter-monomer interactions (hydrogen bonds, ionic, hydrophobic and ring stacking interactions) are calculated with the findbonds option. The interactions here are calculated with the interfacea python package.
The results are saved in the csv format under projectID1/02_interfacea_results & projectID2/02_interfacea_results.
--itypes all calculates hbonds, ionic, hydrophobic, ring_stacking interactions.
You also have the option to calculate intra-monomer interactions with -intra flag.
minnie findbonds -i projectID1 -d 'projectID1/01_frames/' --itypes all --clean
minnie findbonds -i projectID2 -d 'projectID2/01_frames/' --itypes all --clean
In the toy model, the project id folders are kept as defined in the previous step, i.e., sox4 & sox18.
minnie findbonds -i sox4 -d 'sox4/01_frames/' --itypes all --clean
minnie findbonds -i sox18 -d 'sox18/01_frames/' --itypes all --clean
3) The observed interactions are filtered according to a user-defined observation frequency, such that only the interactions that there at least for the x% of the simulation time will be kept. This is achieved with the timefilter option.
minnie timefilter -f projectID1/02_interfacea_results/*/projectID1_merged_*.csv -i projectID1 --per observation_freq
minnie timefilter -f projectID2/02_interfacea_results/*/projectID2_merged_*.csv -i projectID2 --per observation_freq
The toy model interactions, which are observed for at least 25% of the simulation are kept.
minnie timefilter -f sox4/02_interfacea_results/*/sox4_merged_*.csv -i sox4 --per 25
minnie timefilter -f sox18/02_interfacea_results/*/sox18_merged_*.csv -i sox18 --per 25
4) At this step of minnie, common and distinct interaction profiles among the two filtered cases are calculated with the comparecx option.
minnie comparecx -i projectID1 projectID2 --per observation_freq
The 25%-filtered interactions of the toy systems are analyzed.
minnie comparecx -i sox4 sox18 --per 25
5) The common and distinct interaction distribution comparisons are delivered as box-and-whisker plots.
minnie graph -i 'projectID1' 'projectID2' --per 25 --itypes all -b interaction_type -s specific
minnie graph -i 'projectID1' 'projectID2' --per 25 --itypes all -b interaction_type -s common
which translate into the following for our toy model:
minnie graph -i 'sox4' 'sox18' --per 25 --itypes all -b protein-dna -s specific
minnie graph -i 'sox4' 'sox18' --per 25 --itypes all -b protein-dna -s common
All of the example toy-model-related commands are provided in pipeline.sh
6) Aaaand, To visualize minnie results with Pymol
To be run in projectID/04_compare_complex/*_freq_filtered/*freq_perres/complex_specific
sed 's/,/ /g' *_hbonds_compared_*_perres.csv | grep "protein-dna" | grep ":" | awk '{printf "show sticks, (resi %s and chain %s) + (resi %s and chain %s) \n", $6,$2,$7,$3}' | sort -u > pymol_hbonds.pml
sed 's/,/ /g' *_hbonds_compared_*_perres.csv | grep "protein-dna" | grep ":" | awk '{printf "distance i. %s and n. %s and chain %s, i. %s and n. %s and chain %s\n", $6,$8,$2,$7,$9,$3}' | sort -u | sed 's/OP1/O1P/g' | sed 's/OP2/O2P/g' >> pymol_hbonds.pml
Options to visualize each bond type are added to pipeline.sh !!
If you would need to see brief descriptions of the minnie's options
minnie --help
minnie <subcommand> --help
An example of help command - to retrieve usage information of the minnie splitpdbs option
$ minnie splitpdbs --help
Usage: minnie splitpdbs
-i, --id <string> <string>
Project ID for your analyses. One per trajectory(s)
-p, --pdbs [<traj.pdb>] [<traj.pdb>]
Trajectory ID of your complex(s)
Usage example:
minnie splitpdbs -i sox4 sox18 --pdbs sox4.pdb sox18.pdb
minnie splitpdbs -i sox4 --pdbs sox4.pdb
If you would need to remove minnie from your conda setup
conda env remove -n minnie
Also remove the minnie-related lines from your .bashrc
We are grateful to Dr. João Rodrigues (@JoaoRodrigues) for making his interfacea package available for minnie! We also thank Tulay Karakulak (@KarakulakTulay) for her contribution to the conceptualization of this work.