An interpretable artificial intelligence framework for designing synthetic lethality-based anti-cancer combination...

Area of Science:

• Computational Biology
• Artificial Intelligence in Oncology
• Drug Discovery

Background:

• Synthetic lethality (SL) offers a promising avenue for developing synergistic combination therapies in cancer treatment.
• Identifying and mechanistically understanding SL interactions is crucial for advancing cancer therapy design.
• Current artificial intelligence (AI) models for SL prediction often lack interpretability, hindering mechanistic insights.

Purpose of the Study:

• To develop an interpretable AI framework for predicting synthetic lethality (SL) interactions.
• To leverage the interpretable AI framework for the design of SL-based synergistic combination therapies.
• To enhance the mechanistic understanding of SL interactions in cancer.

Main Methods:

• Proposed a knowledge and data dual-driven AI framework for SL prediction, named KDDSL.
• Integrated gene knowledge related to SL mechanisms to guide model construction.
• Developed a method to identify key gene knowledge driving the model's predictions.

Main Results:

• KDDSL demonstrated a favorable balance between predictive accuracy and interpretability.
• Validated findings through experimental and literature-based evidence.
• Successfully identified promising drug combinations, including MDM2 and CDK9 inhibitors, with significant in vitro and in vivo anti-cancer effects.

Conclusions:

• The KDDSL framework shows significant potential for guiding the design of SL-based combination therapies.
• Highlights the need for AI strategies in biomedicine that integrate biological knowledge with predictive models.
• Emphasizes the value of interpretable AI in uncovering therapeutic mechanisms.