scCamAge: A context-aware prediction engine for cellular age, aging-associated bioactivities, and morphometrics

Area of Science:

• Biotechnology
• Computational Biology
• Gerontology

Background:

• Current deep learning models struggle to capture complex spatiotemporal cellular changes during aging.
• Holistic analysis of cellular aging requires integrating diverse data types, including imaging, morphometrics, and molecular bioactivities.

Purpose of the Study:

• To introduce scCamAge, a context-aware multimodal prediction engine for analyzing cellular aging.
• To leverage single-cell image features, morphometrics, and aging-associated bioactivities for comprehensive aging prediction.
• To validate scCamAge's efficacy using yeast models and assess its cross-species applicability.

Main Methods:

• scCamAge integrates image-based cellular spatiotemporal features with morphometrics and aging bioactivities (genomic instability, mitochondrial dysfunction, etc.).
• Utilized heterogeneous datasets of approximately 1 million single yeast cells.
• Validated scCamAge using pro-longevity drugs, genetic mutants, and stress-induced aging models.

Main Results:

• scCamAge accurately predicted pro-longevity responses in yeast under thermal stress, confirmed by omics analyses.
• Trained on yeast, scCamAge demonstrated cross-species predictive power for human fibroblast senescence without additional training.
• Retraining scCamAge on human fibroblast data further enhanced its senescent cell prediction capabilities.

Conclusions:

• scCamAge offers a powerful, multimodal approach to understanding cellular aging.
• The study highlights the evolutionary conservation of aging-related morphometrics.
• scCamAge shows potential for broad applications in aging research and drug discovery across species.