Hierarchical Virtual Screening for Multi-Target Antidepressant Analogs

Hierarchical Virtual Screening for Multi-Target Antidepressant Analogs virtual screening data This dataset contains the main virtual screening results for the study "Computational Prioritization of Multi-Target Antidepressant Analogs via Hierarchical Virtual Screening". The project utilizes a multi-stage in silico pipeline to identify novel polypharmacological candidates for Major Depressive Disorder (MDD). 1. Directory Structure or depression-virtual-screening.tar.gz . ├── docking/ │ ├── receptors/ # Prepared protein structures (PDB/AlphaFold) │ └── results/ # PDBQT files with binding poses and affinities │ ├── Agomelatine/ # Results for analogs based on Agomelatine │ ├── Amineptine/ # Results for analogs based on Amineptine │ └── ... # (Contains 45 parent drug analog groups) ├── qsar/ │ ├── results/ # Prediction CSVs per target and activity mode │ │ ├── DAT/ # Dopamine Transporter │ │ │ ├── AGONIST/ │ │ │ └── ANTAGONIST/ │ │ └── ... # (Covers 18 biological targets) │ └── bambu_qsar_results.csv # Master matrix of QSAR probabilities ├── similarity_search/ │ └── results/ # Tanimoto similarity scores (.txt) per ligand ├── copy_files.sh # Shell script for automated data management └── files.txt # Complete file manifest 2. Directory Structure or qsar-modelling.tar.gz . ├── bambu_datasets/ # Raw bioactivity datasets │ ├── DAT.AGONIST.csv │ ├── DAT.ANTAGONIST.csv │ └── ... # (CSV files for 18 targets) ├── bambu_models/ # Model binaries, preprocessing, and validation │ ├── bambu_filtered.csv │ ├── preprocessing/ # Fitted preprocessor.pkl and train/test splits │ ├── results/ # QSAR prediction outputs per target │ ├── trainning/ # Trained Extra-Trees model.pkl binaries │ └── validation/ # validation.json metrics per target 3. Technical Specifications Ligand Library Initial Library: 14,561 analogs from the ZINC database. Generation Tool: ChemFP (Fingerprint-based similarity). Templates: 45 clinically established antidepressant drugs. Quantitative Structure-Activity Relationship (QSAR) Framework: BAMBU (Automated QSAR pipeline). Algorithm: Extremely Randomized Trees (Extra-Trees). Descriptors: 2048-bit Morgan Fingerprints. Molecular Docking Software: AutoDock Vina. Selection Criteria: Binding energy ≤ -7.0 kcal/mol and Ligand Efficiency ≤ -0.3 kcal/mol. ADMET Filtering Tool: ADMETLab 2.0. Safety Filters: Exclusion of PAINS, hERG inhibitors, and DILI-positive compounds. CNS Requirements: Predicted Blood-Brain Barrier (BBB) permeability and P-gp non-substrate status. 4. Citation and Contact Citation: Höfs, F. N., Lourenço, D. A., Dias, R. S., & Kremer, F. S. (2026). Computational Prioritization of Multi-Target Antidepressant Analogs via Hierarchical Virtual Screening. Corresponding Author: Frederico Schmitt Kremer ([e-mail])