ADMETlab 3.0

ADMETLab3 profiles absorption, distribution, metabolism, excretion, and toxicity across a wider endpoint set than a first-pass screen covers. Reach for it during lead optimization and regulatory-grade profiling — when you need comprehensive, decision-ready pharmacokinetic and safety data on a shortlist of compounds rather than a fast triage pass. For a fast, broad first screen of a large library, start with ADMET-AI and bring the survivors here. See what you can achieve for the full set.

ADMETlab 3.0 integration in .bloom* platform

ADMETlab 3.0 is one of the most popular ADMET prediction platforms, with over 1.7 million visits to its web server. Published in Nucleic Acids Research (2024), this third version provides comprehensive pharmacokinetic and toxicity predictions with 119 prediction endpoints—an increase of 31 compared to the previous version—trained on over 400,000 entries (1.5× larger than ADMETlab 2.0).

Citations

Xiong G, Wu Z, Yi J, Fu L, Yang Z, Hsieh C, Yin M, Zeng X, Wu C, Lu A, Chen X, Hou T, Cao D. ADMETlab 2.0: an integrated online platform for accurate and comprehensive predictions of ADMET properties. Nucleic Acids Research. 2021 Jul 2;49(W1):W5-W14. doi: 10.1093/nar/gkab255

Fu L, Shi S, Yi J, Wang N, He Y, Wu Z, Peng J, Deng Y, Wang W, Wu C, Lyu A, Zeng X, Zhao W, Hou T, Cao D. ADMETlab 3.0: an updated comprehensive online ADMET prediction platform enhanced with broader coverage, improved performance, API functionality and decision support. Nucleic Acids Research. 2024 Jul 5;52(W1):W422-W431. doi: 10.1093/nar/gkae236

Technical foundation

Model architecture

ADMETlab 3.0 incorporates a multi-task DMPNN (Directed Message Passing Neural Network) architecture coupled with molecular descriptors. This approach:

• Guarantees calculation speed for all endpoints simultaneously
• Achieves superior performance in accuracy and robustness
• Provides uncertainty estimates for confident compound selection

Key improvements in version 3.0

• 119 total endpoints (31 new compared to v2.0)
• API functionality for programmatic access to large-scale predictions
• Decision support features for compound prioritization
• Uncertainty quantification in prediction results

Webapp prediction capabilities

Comprehensive ADMET endpoint coverage (119 endpoints)

Absorption predictions (new in v3.0: human oral bioavailability 50%, PAMPA):

• Caco-2 permeability: Intestinal absorption assessment
• PAMPA: Parallel Artificial Membrane Permeability Assay (new)
• Human oral bioavailability: F20% and F50% classification (new)
• P-glycoprotein interaction: Efflux transporter effects
• Solubility: Aqueous solubility estimates
• Lipophilicity: Partition coefficient calculations

Distribution predictions (new: transmembrane transporter inhibitors):

• Plasma protein binding: Fraction unbound calculations
• Blood-brain barrier penetration: CNS exposure assessment
• Volume of distribution: Tissue distribution estimates
• Transporter inhibition: BCRP, OATP1B1, OATP1B3, OCT2 inhibitors (new)

Metabolism predictions (new: CYP inhibition data, microsomal stability):

• CYP450 inhibition: CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP3A4
• Human liver microsomal stability: Metabolic stability assessment (new)
• Metabolic stability: Half-life predictions
• Clearance rates: Hepatic clearance estimates

Excretion predictions:

• Renal clearance: Kidney elimination assessment
• Biliary excretion: Liver excretion pathways
• Half-life: Elimination kinetics
• Bioavailability: Oral absorption estimates

Toxicity predictions (new: nephrotoxicity, neurotoxicity, ototoxicity, hematotoxicity):

• Hepatotoxicity (DILI): Drug-induced liver injury risk
• Cardiotoxicity (hERG): hERG channel blockers
• Nephrotoxicity: Kidney toxicity risk (new)
• Neurotoxicity: CNS toxicity assessment (new)
• Ototoxicity: Hearing damage potential (new)
• Hematotoxicity: Blood cell toxicity (new)
• Genotoxicity: DNA damage potential (new)
• Cytotoxicity: A549, Hek293 cell toxicity (new)
• Immunotoxicity: RPMI-8226 immunotoxicity (new)

Automated compound processing

Structure preprocessing:

• SMILES standardization: Canonical representation
• Salt stripping: Neutral form preparation
• Stereochemistry handling: Isomer-specific predictions
• Tautomer normalization: Consistent protonation states

Quality validation:

• Structure validation: Chemical validity checking
• Molecular weight assessment: Size-appropriate predictions
• Functional group analysis: Reactive moiety identification
• Drug-likeness evaluation: Pharmaceutical compound assessment

Error handling:

• Per-compound feedback: Individual validation results
• Failure isolation: Continue processing despite single failures
• Detailed error reporting: Actionable issue descriptions
• Recovery suggestions: Fix recommendations for failed compounds

When to use ADMET-Lab 3.0 in drug discovery

Comprehensive profiling applications

Lead optimization:

• Evaluate multiple ADMET properties simultaneously
• Identify optimization opportunities across endpoints
• Balance potency gains with safety and PK improvements
• Generate comprehensive risk-benefit profiles

Compound library screening:

• Assess ADMET properties across large compound collections
• Prioritize compounds with favorable profiles
• Flag molecules requiring additional optimization
• Generate decision-ready datasets for project teams

Safety assessment:

• Evaluate toxicity risks early in discovery
• Assess cardiovascular and hepatic safety
• Identify potential drug-drug interaction liabilities
• Generate safety profiles for regulatory submissions

Comparative analysis workflows

Model validation:

• Cross-validate predictions with other ADMET tools
• Compare performance across different model families
• Assess prediction confidence and reliability
• Validate results against experimental data

Series comparison:

• Compare ADMET profiles across analog series
• Track property changes during optimization campaigns
• Identify optimal compounds within series
• Generate SAR analysis with PK context

Webapp interface features

Compound input methods

Direct entry options:

• SMILES input: Single compound or batch entry
• File upload: CSV/SDF files with compound collections
• Previous jobs: Import from generative chemistry tools
• Interactive drawing: Molecular sketcher for custom compounds

Batch processing:

• Support for hundreds of compounds per job
• Progress monitoring with real-time updates
• Error handling for individual compound failures
• Result aggregation and statistical analysis

Result presentation and analysis

Comprehensive output tables:

• All ADMET endpoints in organized columns
• Confidence scores for each prediction
• Risk categories (Low/Medium/High) for safety endpoints
• Molecular properties and descriptors

Interactive visualizations:

• Risk heatmaps: Visual toxicity and safety assessment
• Property distributions: Statistical analysis across datasets
• Correlation plots: Relationships between different endpoints
• Compound comparison: Side-by-side analysis tools

Quality metrics:

• Prediction confidence: Model certainty for each result
• Applicability domain: Assessment of compound suitability
• Model versions: Tracking for reproducibility
• Validation flags: Data quality indicators

Integration with .bloom* ecosystem

Upstream compound sources

Chemistry tools integration:

• DeepPurpose: Evaluate generated molecule properties
• AiZynthFinder: Assess synthetic intermediate profiles
• ZINC/PubChem: Profile database compounds

Research workflows:

• Virtual screening: ADMET evaluation of hit compounds
• Lead optimization: Comprehensive profiling of series
• Compound libraries: Quality assessment of collections
• SAR analysis: PK context for structure-activity relationships

Downstream applications

Decision support:

• Compound advancement: Go/no-go decisions with PK data
• Optimization strategy: Multi-parameter optimization guidance
• Risk assessment: Safety profile generation
• Regulatory preparation: Early ADMET data collection

Experimental planning:

• In vitro assays: Prioritization based on predictions
• In vivo studies: PK parameter optimization guidance
• Formulation development: Solubility and permeability considerations
• Safety testing: Risk-based toxicity study design

Performance and reliability

Prediction accuracy and validation

Model performance:

• Correlation coefficients: >0.7 for most endpoints
• Classification accuracy: >80% for binary predictions
• External validation: Tested against diverse datasets
• Regular updates: Model improvements and recalibration

Confidence assessment:

• Prediction uncertainty: Confidence intervals provided
• Applicability domain: Compound suitability checking
• Model limitations: Known areas of reduced accuracy
• Version tracking: Reproducibility across predictions

Scalability and throughput

Batch processing capabilities:

• Support for large compound libraries
• Parallel processing for improved speed
• Memory optimization for complex molecules
• Queue management for high-throughput needs

Resource optimization:

• Automatic batch sizing for optimal performance
• GPU acceleration where available
• Memory management for large datasets
• Error recovery and retry mechanisms

Best practices for ADMET-Lab 3.0 usage

Input preparation

Compound quality:

• Ensure valid SMILES representations
• Remove counterions and salts where appropriate
• Verify molecular weights are in appropriate ranges
• Check for unusual or reactive functional groups

Batch organization:

• Include compound identifiers for result tracking
• Group similar compounds for consistent analysis
• Consider batch size vs. processing time trade-offs
• Plan for potential individual compound failures

Result interpretation

Risk assessment:

• Use provided risk categories rather than custom thresholds
• Consider confidence scores when making decisions
• Cross-validate important predictions with experimental data
• Account for prediction uncertainty in decision making

Comparative analysis:

• Use consistent parameters when comparing compound series
• Track model versions across different analyses
• Consider applicability domain limitations
• Validate results against known compounds when possible

Workflow integration

Research pipeline:

• Plan ADMET assessment early in compound evaluation
• Use results to guide optimization strategies
• Integrate with other prediction tools for comprehensive profiling
• Generate reports for project team decision making

Quality assurance:

• Document analysis parameters for reproducibility
• Validate key predictions experimentally when possible
• Monitor model performance and updates
• Maintain audit trails for regulatory compliance

Chat Usage

Example Prompts

• "Run ADMETLab3 on these 5 SMILES strings"
• "Fetch gefitinib, erlotinib, and osimertinib from PubChem, then run ADMETLab3 — compare CYP inhibition profiles"
• "Screen my REINVENT hits through ADMETLab3 for hERG and CYP3A4 flags"

Workflow Chaining

• Feeds into: Compound Passport (ADMET section), Safety Passport, AiZynthFinder (compounds passing ADMET)
• Accepts from: PubChem, ZINC, ChEMBL (SMILES), REINVENT/GenMol (generated molecules), previous ADMET jobs

FAQ

Read next