Enhancements in ADMET Predictor for Molecular Predictions

The document is a technical report detailing the enhancements made to ADMET Predictor for predicting properties of beyond Rule-of-Five (bRo5) molecules, including macrocycles and PROTACs. It describes the introduction of new descriptors that capture molecular chameleonicity, which allows for improved predictions of critical high-throughput pharmacokinetics (HTPK) parameters such as liver microsome clearance, fraction unbound in plasma, and blood-to-plasma ratio. The report outlines three specialized models: EPSA1, ChromLogD2, and ChameLogK3, which serve as advanced molecular features. The validation results on novel compounds demonstrate significant performance improvements for bRo5 molecules, showing enhanced accuracy in simulations. Additionally, the report discusses the limitations of Lipinski's Rule of Five regarding drug-like molecules and highlights new training methods for filter models specifically for PROTAC and cyclic peptides. The findings support the ongoing advancements in drug discovery methodologies.