Virtual Screening Pipeline for Potential Tridentate 1,2,4-Triazine Ligands for Cu²⁺ Sensing

An end-to-end cheminformatics workflow for the automated discovery and prioritization of potential tridentate 1,2,4-triazine ligands relevant to Cu²⁺ sensing applications. Starting from a virtual hydrazide library, the pipeline performs reaction enumeration, molecular descriptor generation, SMARTS-based donor atom and chelation motif identification, structural filtering, molecular fingerprint similarity analysis using manually curated tridentate ligand scaffolds, and semiempirical quantum chemical calculations (GFN2-xTB) to prioritize candidates based on their structural and electronic properties. The workflow is fully reproducible, modular, and implemented using open-source Python tools, providing a scalable framework for computational ligand discovery.

---

Top Ranked Candidate Molecules

Top 30 candidate tridentate ligands selected after structural filtering, similarity analysis, and quantum chemical screening.

---

Project Highlights

• End-to-end automated virtual screening workflow
• Virtual reaction enumeration of 1,2,4-triazine derivatives
• SMARTS-based identification of donor atoms and chelation motifs
• Physicochemical descriptor generation using RDKit
• Identification of potential tridentate ligand architectures
• Molecular fingerprint similarity analysis using manually curated reference ligands
• Semiempirical quantum chemical screening using GFN2-xTB
• Automated ranking of candidates based on structural and electronic descriptors
• Modular and reproducible Python workflow

---

Workflow

Hydrazide Library
        │
        ▼
Virtual Reaction Enumeration
        │
        ▼
Molecular Descriptor Generation
        │
        ▼
SMARTS-Based Structural Filtering
        │
        ▼
Donor Atom & Chelation Motif Identification
        │
        ▼
Reference Ligand Similarity Analysis
        │
        ▼
GFN2-xTB Quantum Calculations
        │
        ▼
Candidate Ranking & Visualization

A graphical workflow diagram can be substituted for the above schematic.

---

Screening Summary

Stage	Description	Molecules Remaining
Initial Library	Virtual hydrazide library	10000
Reaction Enumeration	Virtual 1,2,4-triazine derivatives	5941
Structure Validation	RDKit sanitization and validation	4999
Descriptor Calculation	Molecular descriptor generation	4951
SMARTS Filtering	Identification of donor atom-containing molecules	3263
Chelation Screening	Selection of potential tridentate ligand motifs	1214
Similarity Analysis	Comparison with manually curated reference ligands	183
Quantum Screening	GFN2-xTB electronic property calculations	30

---

Reference Ligand Set

The similarity analysis was performed using a manually curated collection of representative tridentate ligand scaffolds identified during structural inspection of the virtual library. These reference compounds were selected based on their donor atom arrangement and chelation geometry rather than experimentally validated Cu²⁺ sensing activity. Morgan fingerprint similarity was subsequently used to identify structurally related candidates exhibiting comparable coordination motifs for further quantum chemical screening.

---

Technologies

• Python
• RDKit
• Pandas
• NumPy
• Matplotlib
• Jupyter Notebook
• GFN2-xTB

---

Repository Structure

TRIAZINE_METAL_SENSING_PROJECT
│
├── data
│   ├── hydrazide_library.csv
│   └── reference_compounds.csv
│
├── notebooks
│   ├── 01_hydrazide_cleaning.ipynb
│   ├── 02_virtual_reaction.ipynb
│   ├── 03_Metal_sensing_App.ipynb
│   └── 04_Quantum_Screening.ipynb
│
├── results
│   ├── figures
│   │   ├── top30_molecules.png
│   │   └── radar_for_top5.png
│   │
│   ├── quantum_screening_data
│   ├── Metal_sensing_mol.csv
│   ├── electronic_prop.csv
│   ├── hydrazides.csv
│   ├── triazines.csv
│   └── xtb_baseline_screening.csv
│
├── src
│   └── automated_screening.py
│
├── cheminfo.yml
├── README.md
└── LICENSE

---

Repository Outputs

Output	Description
triazines.csv	Complete virtual triazine library
Metal_sensing_mol.csv	Filtered candidate tridentate ligands
electronic_prop.csv	HOMO, LUMO and HOMO–LUMO gap values
xtb_baseline_screening.csv	Electronic properties from GFN2-xTB calculations
top30_molecules.png	Top-ranked candidate structures
radar_for_top5.png	Comparison of normalized molecular and electronic properties of the top five candidates

---

Electronic Property Comparison

Normalized comparison of molecular descriptors and electronic properties for the five highest-ranked candidate ligands.

---

Installation

Clone the repository

git clone https://github.com/<username>/TRIAZINE_METAL_SENSING_PROJECT.git
cd TRIAZINE_METAL_SENSING_PROJECT

Create the environment

conda env create -f cheminfo.yml
conda activate cheminfo

Run the complete workflow

python src/automated_screening.py

Alternatively, execute the Jupyter notebooks sequentially to reproduce each stage of the screening pipeline.

---

Future Work

• Density Functional Theory (DFT) optimization of prioritized ligands
• Cu²⁺ complex geometry optimization and binding energy calculations
• Machine learning models for ligand prioritization
• Extension to additional transition-metal ions
• Experimental synthesis and validation of selected candidates

---

License

This project is distributed under the MIT License.