MeetDockOne: a scoring method for protein protein docking

---

Authors:

• Guillaume Delevoye
• François Gravey
• Ilyes Abdelhamid
• Paula Milan Rodriguez
• Maxime Borry

Introduction

This scoring method uses the following steps:

• Checking for shape complementarity and steric clashes.
• Knowledge based scoring on amino-acids interactions.
• Electrostatic and Van der Waals interactions scoring.
• Adjusting the weights of the different scoring functions by machine learning.

Dependancies

• Conda

Download

1- download the latest release of MeetDock here

2- decompress the .zip file

unzip 2017-2018_Equipe1-*

3- go to MeetDockOne folder

cd 2017-2018_Equipe1-*

Installation

A- Environment - Conda

To ensure the cross-platform reproducibility of this method, we choose to work with a conda environment.

The environment file is available here as environment.yml.

1- create the environment from the environment.yml file

conda env create -f environment.yml

2- activate the environment

source activate meetu

3- solvant accessible surface (optional)

The default method to compute the solvant accessible surface is msms.
However, it is really slow and will not work for big complexes, but it will run out of the box and is included in MeetDockOne.

To solve this issue, it is also possible to run MeetDockOne with Naccess (see Documentation section of this readme), a faster and more reliable method.
To do so, you will first need to install Naccess yourself.

The compilation of Naccess requires a Fortran compiler.

Ubuntu/Debian Linux

• sudo apt-get install gfortran
• csh lib/NACCESS/install.scr
• cp lib/NACCESS/* /usr/local/bin

OS X

• Download and install the fortran compiler for your system here
• csh lib/NACCESS/macinstall.scr
• cp lib/NACCESS/* /usr/local/bin

Our two programs

There are two programs you can use from the root directory of the project

meetdock : meetdock is our scoring-only program for the other teams.

meetdockAZ : meetdockAZ is our full program that includes the work of team sampling n°6 and our scoring.

Quick start with meetdockAZ

meetdockAZ is our full implementation from sampling to scoring.

Just type

(meetu)$ python meetdockAZ.py

The program will then ask you if you want to run a demo on a preconfigured ligand+receptor (without minimizer) and very little amount of ligand rotations

If you say no, then it will ask you to manually specify your inputs (receptor, chain, outputdir, minimizer, nb of rotations)

Default output

• By default, output is in ./out/

Quick start with meetdock

(meetu)$ ./meetdock  path/to/meetdock/data -recChain A -ligChain B  -shape -electro -jones -proba

• The dictory path/to/meetdockone/data contains pdb complexes to score.
• The receptor is Chain A (-recChain A).
• The ligand is Chain B (-LigChain B).
• The following methods will be computed on the complexes :
  • shape complementarity (-shape)
  • electrostatic energy (``-electro`)
  • Leenard-Jones interactions (-jones)
  • A knowledge based scoring function (-proba)

Output

• MeetDockOne_results.csv: a .csv file summarizing your results, sorted by decreasing predicted Tm score.
  • pdb: name of the pdb file
  • electrostatic: Electrostatic energy
  • shape: Shape complementarity, similar to Chen et al.
  • vdw: Van der Waals interactions
  • tm_score_prediction: Predicted Tm score with our Machine learning model
  • interpretations: Tm score > 0.8 Excellent || 0.8 > Tm score >= 0.6 Moyen || 0.6 > Tm score >= 0.4 Passable || 0.4 > Tm score mauvais

Example:

pdb	electro	shape	statpot	vdw	tm_score_prediction	interpretations
4LW4.pdb	1,2875610561314053e-25	-15964,0	26,619999999999997	-210,20280214660673	0,5681633333333332	passable
2za4.pdb	1,5077200870763863e-25	-1982,0	29,1	21,831874052394138	0,5578466666666666	passable
4H03.pdb	-3,2388259307234537e-26	-3307,0	19,75	27,762327090226467	0,5353899999999999	passable
2IDO.pdb	-0,0	-4082,0	14,41	-81,43243931391687	0,52383	passable

• MeetDockOne_results.png: a plot of the different scores distributions

Example:

Documentation

(meetu)$ ./meetdock -h
usage: meetdock    [-h] [-shape] [-electro] [-jones] [-proba] [-outdir OUTDIR]
                   [-recChain RECCHAIN] [-ligChain LIGCHAIN] [-depth DEPTH]
                   [-pH PH] [-dist DIST] [-thread THREAD]
                   pdbpath

MeetDockOne scores a protein complex docking

positional arguments:
  pdbpath             absolute path to pdb complex directory. Each pdb file
                      must contain the receptor and the ligand as separated
                      chains.

optional arguments:
  -h, --help          show this help message and exit
  -shape              compute shape complementarity
  -electro            compute Electrostatic interactions
  -jones              compute Lennard-Jones interactions
  -proba              Compute knowledge based interactions
  -outdir OUTDIR      path to (existing) output directory
  -recChain RECCHAIN  receptor Chain ID. If there more than one, separate by a
                      ','. Default = 'A'
  -ligChain LIGCHAIN  ligand Chain ID. If there more than one, separate by a
                      ','. Default = 'B'
  -depth DEPTH        Method for residue depth/solvant accessibility. [msms |
                      naccess]. Default = msms
  -pH PH              pH for electrostatic interactions. Default = 7
  -dist DIST          Threshold for interface determination (Angstrom).
                      Default = 8.5
  -thread THREAD      Number of threads for MultiThreading. Default: automatic
                      detection