MegaEye: Applying multiple machine learning approaches to identify oral compounds with ocular bioactivity.

Area of Science:

• Ophthalmology and Computational Chemistry
• Drug Discovery and Development
• Biomedical Informatics

Background:

• The eye possesses physiological barriers hindering oral drug delivery for ocular diseases.
• Predicting a molecule's ability to cross ocular barriers is crucial for treatment efficacy.
• Machine learning applications for predicting ocular drug bioactivity are currently underexplored.

Purpose of the Study:

• To curate datasets and develop machine learning models for predicting ocular drug penetration after oral administration.
• To identify potential drug candidates for treating ocular diseases via oral delivery.
• To explore the utility of large language models in ranking natural compounds for ocular activity.

Main Methods:

• Curation of datasets including molecular properties, blood-brain barrier MPO scores, and blood-retinal barrier proxies.
• Development and validation of machine learning models using FDA-approved drugs with reported ocular activity.
• Application of a large language model to rank over 400,000 natural compounds for potential ocular activity.

Main Results:

• Successfully curated diverse datasets to train predictive models for ocular drug delivery.
• Validated machine learning models' capability to identify molecules with potential ocular activity.
• Identified and ranked a large set of natural compounds based on their predicted ocular bioactivity.

Conclusions:

• Machine learning offers a promising avenue for predicting ocular drug penetration and identifying novel therapeutic candidates.
• The developed models and approach can be expanded for broader ocular applications, including drug repurposing.
• This study demonstrates the potential of computational methods to accelerate the discovery of treatments for eye conditions.