Machine learning for compound potency prediction

--------------------- MACHINE and DEEP LEARNING SCRIPTS: GENERAL INFORMATION ---------------------

This folder contains scripts to build and analyse ML models.

Should be downloaded and all Jupyter notebooks can be used inside the folder.

The content of the folders is summarized in the following.

Jupyter Notebooks (.ipynb):

(1) ml_models: Jupyter notebook with machine and deep learning models (MR, kNN, SVR, RFR, DNN), able to predict compound potency for different datasets (Complete/Random/Diverse sets)

(2) gcn_models: Jupyter notebook with graph neural networks models (GCN), to predict compound potency for different datasets (Complete/Random/Diverse sets)

(3) ml_models_cluster_potent_sets: Jupyter notebook with machine and deep learning models (MR, kNN, SVR, RFR, DNN), able to predict compound potency for different datasets (Cluster/Potent sets)

(4) gcn_models_cluster_potent_sets: Jupyter notebook with graph neural networks models (GCN), able to predict compound potency for different datasets (Cluster/Potent sets)

(5) data_analysis_figures: Jupyter notebook with a workflow for the data analysis of compound potency predictions from regression models.

Python scripts (.py):

(6) ml_utils: script that provide supporting functions for ML/DL models generation

(7) fingerprint: script to calculate molecular fingerprints (Morgan fingerprints)

(8) machine_learning_models: script to build ML/DL models for regression (MR, kNN, SVR, RFR, DNN)

(9) Folders:

- dataset: stores the compound potency dataset used in this analysis
	- dataset_analysis: Jupyter notebook for the dataset analysis

- ccr_results : stores the CCR algorithm results

- regression_results: stores regression models predictions

(10) Python environment:

- ML_env.yml provides the python environment used for this analysis. (Requires instalation see below)

Order of Jupyter notebook execution:

1. (1), (2), (3), (4) to generate the model predictions (results).

2. (5) to generate the figures.

Python environment installation:

1. Open Anaconda command line

2. Type 'conda env create -n ENVNAME --file ENV.yml', where 'ENVNAME' is the desired environment and 'ENV' the full path to the yml file.

Python environment export:

1. Open Anaconda command line

2. Type 'conda env export ENVNAME>ENV.yml', where 'ENVNAME' is the desired environment and 'ENV' the full path to the yml file.

Requirements:

• python=3.9
• scipy=1.8.1
• numpy=1.22.4
• scikit-learn==1.1.1
• tensorflow==2.9.1
• keras==2.9.0
• rdkit==2022.3.3
• cudatoolkit=11.2.2
• dgl-cuda11.1=0.8.1
• deepchem==2.6.1
• tqdm=4.64.0