Predicting oral druglikeness by iterative stochastic elimination

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and development
• Pharmacokinetics and drug bioavailability

Background:

• Computational methods are crucial for early-stage drug discovery to improve success rates.
• High-quality drug candidates are essential to minimize clinical attrition.
• Predicting oral drug likeness is a key challenge in pharmaceutical research.

Purpose of the Study:

• To present a novel computational approach for indexing the oral drug likeness of compounds.
• To introduce the Iterative Stochastic Elimination (ISE) Algorithm for distinguishing orally available drugs.
• To develop an Orally Bioavailable Druglike Index (OB-DLI) for prioritizing drug candidates.

Main Methods:

• Utilized the Iterative Stochastic Elimination (ISE) Algorithm to generate optimized descriptor ranges as 'filters'.
• Developed a 'filter bank' by clustering diverse k-descriptor sets.
• Combined filters into an Orally Bioavailable Druglike Index (OB-DLI) for molecular characterization.

Main Results:

• A single filter with 5 descriptors achieved 81% true positives and >77% true negatives.
• The OB-DLI demonstrated higher discriminative power and improved ranking of oral drug candidates compared to binary decisions.
• The ISE approach identified structurally dissimilar molecules with desired oral bioavailability properties.

Conclusions:

• The OB-DLI offers a robust method for prioritizing molecules based on oral drug likeness.
• This computational approach significantly improves prediction accuracy and aids in discovering novel drug candidates.
• The ISE algorithm and OB-DLI show a >13% improvement in Matthews Correlation Coefficient over existing methods.