Enhancing ERα-targeted compound efficacy in breast cancer threapy with ExplainableAI and GeneticAlgorithm

Area of Science:

• Computational chemistry and cheminformatics
• Artificial intelligence in drug discovery
• Oncology therapeutics

Background:

• Breast cancer is a leading cause of female mortality worldwide.
• There is a critical need for advanced therapeutic strategies.
• Targeting estrogen receptor alpha is a key approach in breast cancer treatment.

Purpose of the Study:

• To develop a comprehensive methodology for enhancing the bioactivity and ADMET properties of compounds targeting estrogen receptor alpha.
• To integrate explainable AI, machine learning, and genetic algorithms for drug discovery.
• To identify and refine critical molecular descriptors for predicting compound efficacy.

Main Methods:

• Employed SHAP (SHapley Additive exPlanations) and LassoNet to identify key molecular descriptors.
• Utilized genetic algorithms for optimizing candidate compounds.
• Validated selected descriptors and predictive models for bioactivity and ADMET properties.

Main Results:

• Identified 50 critical molecular descriptors from 729, significantly influencing bioactivity prediction.
• Achieved a mean R-squared of 77% for bioactivity prediction.
• Obtained high accuracy scores for ADMET properties: 90.2% (Absorption), 93.7% (Distribution), 89.5% (Metabolism), 87.3% (Excretion), and 95.8% (Toxicity).
• Optimized compounds demonstrated superior bioactivity with pIC50 values up to 10.05.

Conclusions:

• The integrated AI and machine learning approach effectively enhances compound bioactivity and ADMET profiles.
• This methodology accelerates the discovery of novel and effective breast cancer therapies.
• The findings highlight the potential of advanced computational techniques in drug development for estrogen receptor alpha-positive cancers.