A common challenge in drug discovery and lead optimization is to proactively identify and avoid potential liabilities of compounds before synthesis. Our Eurofins Discovery Research Informatics team has developed a suite of computational tools to evaluate the likelihood of compounds having non-ideal ADME properties during compound lead optimization, by leveraging AI-methods and data from the industry-leading BioPrint database (designed to assess deeper mechanistic understanding of safety liabilities and to search for compounds with similar profiles). Our models predict the likelihood that a given compound will have reasonable Solubility, Permeability, Plasma Protein Binding prediction, Metabolic Stability prediction, Efflux prediction, CYP prediction (whole list) and hERG prediction properties. Our approach has matched the performance of most large pharmaceutical companies by combining our proprietary data sets and ML tools.
| Attribute | State of the Art Predictor | State of the Art Performance (within AD*) |
RI-CLO Predictors (within AD*) |
|---|---|---|---|
| Solubility | Bayer, JNJ | MCC > 0.4 | MCC 0.59 |
| Permeability | Bayer, JNJ | MCC > 0.4 | MCC 0.44 |
| Efflux Ratio | Bayer, JNJ | MCC > 0.4 | MCC 0.46 |
| Metabolic Stability | JNJ | MCC > 0.4 | MCC 0.43 |
| Plasma Protein Binding | JNJ | MCC > 0.4 | MCC 0.59 |
| hERG | Bayer | MCC > 0.4 | MCC 0.31* |
| 1A2 | JNJ | MCC > 0.4 | MCC 0.62 |
| 2D6 | JNJ | MCC > 0.4 | MCC 0.50 |
| 2C9 | JNJ | MCC > 0.4 | MCC 0.47 |
| 2C19 | JNJ | MCC > 0.4 | MCC 0.50 |
| 3A4 | JNJ | MCC > 0.4 | MCC 0.47 |
This suite of high-performance prediction models can be used at any stage of the drug discovery process from Hit Identification, Lead Identification, to full integration in Lead Optimization. And the technology used to develop these predictors can be used to develop local models to support efficient progress of integrated drug discovery projects.
