Random forest classification for predicting lifespan-extending chemical compounds.

Area of Science:

• Gerontology and computational chemistry.
• Biogerontology and drug discovery.

Background:

• Aging is a significant risk factor for numerous diseases, including cancer, cardiovascular, and neurodegenerative conditions.
• Pharmaceutical interventions targeting aging and age-related diseases are a rapidly expanding research domain.

Purpose of the Study:

• To construct a machine learning model utilizing the DrugAge database to predict compounds that extend the lifespan of Caenorhabditis elegans.
• To identify key molecular features predictive of lifespan extension in chemical compounds.

Main Methods:

• Developed five predictive models using the random forest algorithm with molecular fingerprints and/or descriptors.
• Utilized Gini importance to rank the top 30 predictive features, focusing on atom/bond counts and topological/partial charge properties.
• Applied the best-performing model (AUC 0.815) to screen an external database of 1738 small molecules.

Main Results:

• The top-performing model, based on molecular descriptors, achieved an Area Under the Curve (AUC) of 0.815.
• Key predictive features included descriptors related to atom and bond counts, and topological and partial charge properties.
• Screening identified flavonoids, fatty acids and conjugates, and organooxygen compounds as potential lifespan-extending agents (predictive probability ≥ 0.80).

Conclusions:

• Machine learning models can effectively predict lifespan-extending compounds for Caenorhabditis elegans.
• Molecular descriptors provide valuable insights into the chemical properties associated with lifespan extension.
• The identified compound classes offer promising avenues for future anti-aging drug discovery research.