Evaluation of ADMET Predictor as a design, prioritization and ‘Tier zero’ screening tool in early drug discovery
Poster Authors:
John Pollard, Michael Bestwick, Elnaz Gozalpour, David Lugo, Simon Teague, Simon Taylor
Pharmaron, West Hill Innovation Park, Hertford Road, Hoddesdon, Hertfordshire, EN11 9FH, United Kingdom
Streamlining Discovery with In Silico Tools and Machine Learning
The increasing pace and complexity of early-stage drug development demand tools that deliver quick, data-driven insights. ADMET Predictor by Simulations Plus is one such solution: a powerful in silico modeling platform that leverages machine learning to predict over 175 ADMET-related properties directly from molecular structure.
Used as a “Tier Zero” screening tool, ADMET Predictor enables discovery teams to triage thousands of compounds virtually before synthesis or in vivo testing. Pharmaron evaluated the latest version (v12) of this tool to assess its effectiveness in early design, prioritization, and dose prediction workflows.
Why Use ADMET Predictor?
1. Structure-Based Screening
Predicts human-relevant PK and ADME properties such as:
- LogD, LogP, protein binding (Fu), clearance
- Intrinsic clearance in hepatocytes and microsomes
- Solubility, permeability, fraction absorbed (%Fa)
Capabilities
2. Reduce Animal Use & Lab Burden
Helps limit experimental testing to only the most promising candidates. Aligns with FDA and industry goals for animal reduction.
3. Real-World Validation
In a retrospective study:
- 113 marketed drugs tested
- R² = 0.9 for dose predictions
- 82% of predictions within 3-fold of observed data
Applications in Early-Phase Programs
Virtual Screening of Project X
Pharmaron applied the Predictor to 790 compounds across three series. The tool effectively:
- Ranked compounds by dose-limiting properties
- Identified a candidate with low clearance and high potency
- Prioritized leads with predicted oral doses <100 mg
DMTA Integration
The Predictor was used to:
- Flag ADME liabilities before synthesis
- Guide compound prioritization in design-make-test-analyze cycles
- Estimate human dose from biochemical potency and PK predictions
Delivering Predictive Insight at Design Time
ADMET software and prediction tools help researchers model complex drug behavior early. It supports smarter chemistry decisions, tighter resource management, and fewer downstream failures. By combining machine learning, curated datasets, and in silico modeling, it accelerates discovery while reducing cost and ethical burden.
References:
- Simulations Plus Official Site
- FDA Guidance on Animal Testing Alternatives
- Machine Learning in Drug Discovery (Nature Review)
Read the poster to explore complete data from Pharmaron’s ADMET Predictor evaluation.