ADMET Prediction - Usage Guide

Overview

Predict absorption, distribution, metabolism, excretion, and toxicity properties for drug discovery. Includes drug-likeness filtering and structural alerts.

Prerequisites

pip install rdkit requests
pip install deepchem  # For ML-based predictions

Quick Start

Tell your AI agent what you want to do: - "Predict ADMET properties for my lead compounds" - "Filter my library for drug-like compounds" - "Check for PAINS alerts in my hit list" - "Predict hERG liability for my compounds"

Example Prompts

ADMET Prediction

"Use ADMETlab 3.0 to predict ADMET properties for these SMILES."

"Predict CYP inhibition profiles for my lead series."

Drug-Likeness

"Calculate Lipinski violations and QED scores for my compounds."

"Annotate Lipinski and Veber results, then rank using thresholds justified for this project."

Safety Filtering

"Check my hits for PAINS and other structural alerts."

"Identify compounds with potential hERG liability."

What the Agent Will Do

Calculate drug-likeness properties (Lipinski, QED)
Use the current official ADMETlab 3.0 API workflow or web service for predictions
Flag PAINS and structural alerts for review
Rank compounds only after calibrating model outputs and project decision rules
Generate summary report

Tips

ADMETlab 3.0 reports 119 platform features: 77 prediction models, 34 computed properties, and 8 rules
SwissADME has no public API and its terms restrict automated data retrieval; use its documented manual batches
Chemprop ensembles require explicit prediction-time uncertainty estimation and separate calibration when calibrated outputs are needed
DeepChem supports both PyTorch and TensorFlow (TF not deprecated)
QED is a ranking descriptor, not a universal pass/fail definition of drug-likeness
Interpret hERG potency relative to exposure, assay conditions, and safety margin; no single IC50 cutoff establishes safety
PAINS matches flag possible assay interference for orthogonal testing; they are not categorical exclusions

molecular-descriptors - Calculate descriptors for ML
substructure-search - Filter reactive groups
virtual-screening - Screen after ADMET filtering