Synergistic Machine Learning Accelerated Discovery of Nanoporous Inorganic Crystals as Non-Absorbable Oral Drugs

Area of Science:

• Materials Science
• Computational Chemistry
• Pharmacology

Background:

• Machine learning (ML) requires extensive high-quality data for effective drug discovery.
• Non-absorbable oral drugs (NODs) offer safety benefits for chronic diseases due to minimal systemic exposure.
• Empirical discovery of NODs is currently resource-intensive and time-consuming.

Purpose of the Study:

• To develop a synergistic ML method for efficient identification of novel non-absorbable oral drugs (NODs).
• To discover superior NODs with high selectivity, capacity, and stability from inorganic materials.
• To address the unmet clinical need for safe and effective hyperkalemia treatments.

Main Methods:

• Integration of small data-driven multi-layer unsupervised learning.
• In silico quantum-mechanical computations for material property prediction.
• Minimal wet-lab experiments for validation and discovery.

Main Results:

• A novel NH4-form nanoporous zeolite with a merlinoite (MER) framework (NH4-MER) was identified.
• NH4-MER demonstrated superior safety and efficacy in reducing blood potassium (K+) in three animal models.
• The discovered material effectively lowered K+ without releasing sodium (Na+), a critical advantage.

Conclusions:

• The developed synergistic ML method significantly accelerates the discovery of non-absorbable oral drugs (NODs).
• NH4-MER represents a promising therapeutic candidate for hyperkalemia in patients with chronic kidney disease and Gordon's syndrome.
• This approach can be extended to accelerate the discovery of other shape-selective materials.