Anti-senescent drug screening by deep learning-based morphology senescence scoring

Area of Science:

• Biotechnology
• Computational Biology
• Gerontology

Background:

• Cellular senescence is a key factor in aging and age-related diseases.
• Identifying senescent cells is crucial for understanding pathogenesis and developing therapies.
• Senescent cells exhibit distinct morphological characteristics that can be visually identified.

Purpose of the Study:

• To develop a morphology-based deep learning system for identifying senescent cells.
• To create a quantitative scoring system for endothelial cell senescence using convolutional neural networks (CNN).
• To utilize this system for high-throughput screening of anti-senescent drugs.

Main Methods:

• Development of a pre-trained CNN optimized for cellular senescence classification.
• Implementation of a morphology-based CNN system named Deep Learning-Based Senescence Scoring System by Morphology (Deep-SeSMo).
• Application of Deep-SeSMo for evaluating anti-senescent reagents and screening a kinase inhibitor library.

Main Results:

• Deep-SeSMo accurately identified senescent cells based on morphology.
• The system effectively evaluated known anti-senescent reagents.
• Four novel anti-senescent drugs were identified through Deep-SeSMo-based screening.
• RNA sequencing indicated these drugs suppress senescence by inhibiting inflammatory pathways.

Conclusions:

• Morphology-based CNNs are powerful tools for identifying cellular senescence.
• Deep-SeSMo provides a quantitative and efficient method for evaluating senescence.
• This approach facilitates the discovery of novel anti-senescent compounds for therapeutic applications.