Supplementary Material Ecotoxicity parameters for reshaping the cheminformatic dataset filtering in the drug discovery pipeline
. ├── 200625_XGB_5Epo/ │ ├── 200625_XGB_5epo_ML.pkl │ ├── Descriptor-Means.txt │ ├── Descriptor-STDs.txt │ ├── XGB_model_metrics.json │ ├── ROC_curve.png │ ├── PR_curve.png │ ├── confusion_matrix.png │ ├── feature importance files │ └── validation datasets │ ├── 1_OHD_Paper_2_Descriptors_calculator.py ├── 2_OHD_paper_2_Descriptor_analysis.py ├── 3_OHD_paper_2_Clustering_analysis.py ├── 4_OHD_paper_2_Model_use.py ├── 5_OHD_paper_2_Prep_dati_GDS.py ├── 6_OHD_paper_2_Admetscore.py ├── 7_OHD_paper_2_admetscoreparametrico.py ├── 8_OHD_paper_2_Ecoscore.py ├── 9_OHD_paper_2_DLscore.py ├── 10_OHD_paper_2_GDS.py ├── 11_OHD_paper_2_scaffold_analysis.py │ ├── Supplementary_File_*.csv / *.xlsx └── README.md
GENERAL DESCRIPTION
This repository contains the complete computational workflow employed for the classification, prioritization, and scaffold analysis of antiparasitic compounds under a One Health perspective. The pipeline integrates molecular descriptor calculation, exploratory chemoinformatics, machine-learning–based safety classification, prediction-based ADMET and ecotoxicological profiling, and multi-parametric compound ranking through the GreenDrugScore (GDS).
All scripts are provided in execution order and were used to generate the results reported in the manuscript and Supplementary Tables.
MACHINE LEARNING MODEL FILES
Folder: 200625_XGB_5Epo/
This directory contains the pretrained XGBoost classification model used for SAFE/UNSAFE prediction, together with all auxiliary files required for model inference and validation.
The folder includes the serialized model file (.pkl), descriptor normalization parameters (means and standard deviations), feature importance rankings, validation datasets, and model performance metrics and plots, including ROC curves, precision–recall curves, and confusion matrices.
The model contained in this folder is directly accessed by the script 4_OHD_paper_2_Model_use.py. No retraining or modification of the model files is required for standard use.
DESCRIPTION OF SCRIPTS
Scripts are numbered according to their execution order.
DESCRIPTOR CALCULATION
Script: 1_OHD_Paper_2_Descriptors_calculator.py
This script calculates molecular descriptors starting from SMILES representations. In the context of the published work, molecular descriptors were already calculated and provided in Supplementary File 2. Therefore, this script is included for completeness and reproducibility but does not need to be executed to reproduce the reported results.
DESCRIPTOR AND CHEMICAL SPACE ANALYSIS
Scripts: 2_OHD_paper_2_Descriptor_analysis.py 3_OHD_paper_2_Clustering_analysis.py
These scripts perform exploratory chemoinformatics analyses of the compound library, including statistical characterization of molecular descriptors and unsupervised clustering. The analyses are used to assess chemical space distribution, molecular diversity, and structural organization of the dataset.
MACHINE LEARNING–BASED SAFE/UNSAFE CLASSIFICATION
Script: 4_OHD_paper_2_Model_use.py
This script applies the pretrained XGBoost model stored in the 200625_XGB_5Epo/ folder to the full compound library.
Input: Supplementary File 2 containing the full library with molecular descriptors. Output: SAFE/UNSAFE classification labels and associated prediction probabilities.
Only compounds classified as SAFE are retained for subsequent ADMET, ecotoxicological, and drug-likeness analyses.
EXTERNAL ADMET AND ECOTOXICOLOGICAL PREDICTIONS
For compounds classified as SAFE, prediction of physicochemical, ADMET, and ecotoxicological endpoints is performed using external web-based tools, namely ADMETlab 3.0 and ADMETSAR 3.0.
From ADMETSAR 3.0, the endpoints OCT2 and AOT_c are extracted and merged with the ADMETlab 3.0 output to generate a unified prediction file used for score calculation.
DATA PREPARATION FOR SCORE CALCULATION
Script: 5_OHD_paper_2_Prep_dati_GDS.py
This script harmonizes and preprocesses the merged prediction dataset in preparation for score calculation. The procedure includes endpoint normalization, thresholding, binarization or semi-binarization where applicable, and formatting of the input required for ADMETscore, ECOscore, and DLscore computation.
SCORE CALCULATION
The following scripts are executed sequentially.
Script: 6_OHD_paper_2_Admetscore.py This script calculates the ADMETscore.
Script: 7_OHD_paper_2_admetscoreparametrico.py This script calculates the weighted or parametric variant of the ADMETscore.
Script: 8_OHD_paper_2_Ecoscore.py This script calculates the ECOscore based on predicted ecotoxicological endpoints.
Script: 9_OHD_paper_2_DLscore.py This script calculates the drug-likeness score (DLscore).
Each script produces an intermediate output that is subsequently merged for final compound ranking.
GREENDRUGSCORE INTEGRATION
Script: 10_OHD_paper_2_GDS.py
This script integrates ADMETscore, ECOscore, and DLscore into the final GreenDrugScore (GDS), which is used for compound prioritization and ranking in the study.
SCAFFOLD NETWORK ANALYSIS
Script: 11_OHD_paper_2_scaffold_analysis.py
This script performs scaffold-based analysis of prioritized compounds. Murcko scaffolds are extracted and organized into a scaffold similarity network, enabling topological and structural interpretation of the prioritized chemical space.
SUPPLEMENTARY DATA FILES
The repository includes all Supplementary Files referenced in the manuscript, including training and validation datasets, SAFE and UNSAFE compound libraries, descriptor tables, GDS-ranked compound lists, scaffold networks, and downgraded compound sets.
These files allow full reproduction of the computational analyses reported in the study.
REPRODUCIBILITY NOTES
All scripts are written in Python and were executed in a standard scientific Python environment. External predictions from ADMETlab 3.0 and ADMETSAR 3.0 must be generated manually and integrated as described. Script numbering reflects the mandatory execution order required to reproduce the published results.