Towards a unified framework for single-cell -omics-based disease prediction through AI

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Single-cell omics reveals cellular heterogeneity crucial for understanding health and disease.
• Artificial intelligence (AI) offers advanced pattern recognition and predictive modeling for clinical applications.

Discussion:

• The proposed scDisPreAI framework integrates single-cell omics data using AI for disease and stage prediction.
• It establishes a standardized database and employs rigorous preprocessing for reliable biological insights.
• Machine learning and deep learning models are trained for multi-task classification, identifying disease identity and stage.

Key Insights:

• Interpretability techniques (SHAP, attention weights) identify influential genes, serving as potential biomarkers.
• Biomarkers may be shared across different diseases or disease stages, offering broad therapeutic potential.
• scDisPreAI consolidates prediction with biomarker discovery for clinical decision support.

Outlook:

• Future directions include integrating multi-omics data and standardizing protocols for broader applicability.
• Prospective clinical validation is essential to realize the full potential of single-cell AI in precision medicine.
• The framework aims to guide tailored treatments and identify therapeutic targets for drug repurposing.