alexismolinamr/peleai

Create scoring functions from simulation data.

Table of contents

• PeleAI
• Requirements
• Usage

PeleAI

The goal of PeleAI-3D is to create target-specific scoring functions from simulation data obtained from molecular simulations methodologies such as Monte-Carlo or docking.

To do so an graph-based topological description of the binding site containing the ligand is computed, which serves later for fitting a model that predicts either the activitiy or the binding energy of the given pose.

Requirements

Install the conda version of rdkit and activate the evironment before installing the required packages:

conda create -c conda-forge -n PeleAI rdkit
conda activate PeleAI

Then, you can install:

tensorflow
numpy==1.19.0
pandas==1.1.4
prody==1.11
sklearn
xgboost
lightgbm
pyyaml
seaborn

pip install -r requirements.txt

Usage

The methodology can be run in three different fashions: graph, fitting or pipeline.

The first one, graph, is dedicated to generate statistics out from a graph topology of the ligand-binding site complex without fitting the data in any model. The second one, fitting, will provide several options to use the graph based statistics previously generated to fit either a regression or a classification task. The last one, pipeline, provides end-to-end results out of the given input poses. It executes graph, whose output is passed to fitting which will report the model results.

Generate graph statistics (graph)

Parameters

#control file for PeleAI3D - Graph statistics

path: 
output: 
run_name: 
ligand_name: 
selection_radius: 
center: 
decay_function:
nodes: 

Fitting graph statistics (fitting_ffnn)

Parameters

#control file for PeleAI3D - Fitting a Feed Forward Neural Nerwork

path_graph: /path/to/folder/with/pdb/from/profile/
output: /path/to/output/
seed: 42
task: regression
cpus: -1
algorithm: ffnn
learning_rate: 0.0001
epochs: 1000
batch_size: 64
pelePrep: profile

Fitting graph statistics (fitting)

Parameters

#control file for PeleAI3D - Fitting a model

path_graph: 
output: 
test_size:
seed: 
task: 
cpus: 
scaler: 
algorithm: 

Pipelining (pipeline)

Parameters

#control file for PeleAI3D - Pipeline
path: 
output: 
run_name: 
ligand_name: 
selection_radius: 
center: 
decay_function:
nodes: 
# ------------------------------------
#pipe: True
#target: 
# ------------------------------------
test_size: 
seed: 
task: 
cpus: 
scaler: 
algorithm: 

Execution

You may pass the input.yaml file to the peleAI3d.py as follows:

python3 peleAI3d.py input.yaml

The results will be written to the folder indicated in the output argument of the input file.