🔒 AI-powered drug discovery toolkit • NCE design • ADMET & statistics

RADGenT: From targets to new chemical entities

RADGenT is a lightweight research platform designed to create and evaluate new chemical entities (NCEs) that can ultimately progress toward safe, effective new drugs.

The goal is simple: help researchers go from target identification → candidate molecules → in-silico evaluation → data-driven decisions, using modular tools for statistics, ADMET prediction, docking, molecular libraries, machine learning, and environmental toxicity.

Explore tools View drug development steps

🧪 Quick run: Docking + ADMET (front-end preview)

Protein target

Example: 1HVR (HIV-1 protease)

Ligand

Example: aspirin.smi

Scoring

Vina + cavity-aware

ADMET

hERG / CYP / logS / HIA

✔️ This is a UI preview. Connect your own back-end / APIs to run real docking & ADMET pipelines.

🤝 Join the RADGenT Research Community

Help build the future of AI-powered drug discovery. We welcome collaborators from science, medicine, computation and design.

RADGenT tool suite

RADGenT is organised as modular apps. Each app focuses on a specific step of the drug discovery and development workflow. More modules can be added over time.

Available now

📊 STMS — Statistics Tool for Medical Students

STMS provides classical and advanced biostatistics for medical research: t-tests, ANOVA, Chi-square, regression, survival analysis, ROC curves, sample-size calculations and reporting helpers.

Clean interface for clinical trials and observational studies.
Supports group comparisons, correlation, regression, survival analysis.
Exports publication-ready tables and figures.

🚀 Launch STMS

Where STMS fits in drug development

Designing and analysing in-vitro / in-vivo experiments.
Analysing toxicology, efficacy and PK data from preclinical studies.
Supporting early-phase clinical data summaries.

Available now

🧪 RADPT — Rajpal Agrawal Drug Prediction Tool (ADMET)

RADPT takes candidate molecules (Name + SMILES) and predicts key drug-likeness and ADMET properties, helping you filter out poor candidates early.

Basic physicochemical descriptors: MW, LogP, HBD, HBA, TPSA, RB.
Rule-based evaluation (Lipinski, Veber, etc.).
Similarity scoring to reference / known drugs.

🧪 Open RADPT section

Where RADPT fits in drug development

Filtering virtual hits for drug-likeness and developability.
Prioritising molecules before expensive synthesis / in-vitro testing.
Supporting lead optimisation decisions.

Planned module

⚓ RADock — Molecular Docking & Pose Analysis

RADock is a planned module for running and visualising docking workflows: AutoDock Vina, cavity-aware scoring, pose ranking and quick pose inspection.

Upload protein (PDB) and ligands (SMILES / SDF / MOL2 / PDBQT).
Configure docking box, constraints and scoring schemes.
Export ranked poses and scores for downstream analysis.

Planned usage

RADock will feed docking scores and pose-level information into RADPT and ML models, closing the loop between structure-based design and data-driven filtering.

Concept stage

📚 RAlib — Library of Candidate Molecules

RAlib is envisioned as a curated library of designed and screened candidates: store, tag and retrieve molecules along with their docking, ADMET and ML scores.

Central place to manage virtual libraries per project.
Track scores from RADock, RADPT and ML models.
Support decisions like “which series do we advance?”

Planned features

Search by substructure / similarity / project tags.
Export subsets for synthesis or external screening.

Concept stage

🤖 ML Models — QSAR & property prediction

This module will host machine learning models (QSAR, deep learning, graph networks) to predict potency, selectivity, ADMET and other properties from molecular structure.

Train / load ML models on project-specific datasets.
Batch-score virtual libraries from RAlib.
Integrate with RADPT and RADock outputs.

Long-term vision

Combine structure-based methods (docking) with data-driven models to prioritise NCEs with the best balance of affinity, safety and developability.

Future module

🌱 RAEco — Environmental Toxicity of Drugs

RAEco is planned as a module to estimate environmental impact and ecotoxicity of drug candidates and their metabolites (e.g. aquatic toxicity, persistence).

Predict environmental fate (persistence, bioaccumulation risk).
Flag molecules with high predicted ecological hazard.
Support greener by design decision-making.

Why RAEco matters

New drugs should not only be safe for patients but also for the environment. RAEco aims to bring environmental toxicity into early discovery decisions.

Key platform features

Cavity-aware docking

Pocket detection and constraint-guided scoring alongside Vina-style terms.

ADMET predictions

Supervised models for ADME/Tox with clear descriptors.

GPU optional

Scale docking batches on CPU or GPU.

Transparent & auditable

Versioned datasets, model cards, exportable reports.

Drug development journey: where RADGenT fits

1. Target identification & validation

Choose a disease mechanism and identify biological targets (receptors, enzymes, transporters). Validate them in-vitro/in-vivo.

2. Hit discovery & NCE design

Use docking, virtual screening and library design (RADock, RAlib, ML models) to propose new chemical entities that interact with the target.

3. Lead optimisation & preclinical

Optimise potency, selectivity and ADMET (RADPT, ML models), check toxicity and PK, and analyse preclinical data using STMS.

4. Clinical trials & approvals

Design and analyse clinical studies (with biostatistics tools like STMS) and build the evidence package required for regulatory approval.

Try the live docking + ADMET demo (mock)

Ligand docking + ADMET (mock)

Ligand file or SMILES

Target PDB ID Docking engine

How it works

Upload a ligand (SMILES/SDF) and choose a target (PDB ID).
Select docking engine and pocket options.
Press Run to simulate (mocked here) and preview results.

This demo runs locally. Connect your server API to enable real jobs and persistent results.

Bring your molecules. We’ll handle the rest.