Data modeling the interplay between single-cell shape, single-cell protein expression, and tissue state

Area of Science:

• Biophysics
• Cell Biology
• Computational Biology

Background:

• Cell shape is critical for tissue function, but its relationship with protein expression is poorly understood.
• Existing methods struggle to link single-cell morphology to molecular profiles in complex tissues.

Purpose of the Study:

• To investigate the link between cell shape and protein expression in heterogeneous human tissues.
• To leverage multiplexed imaging and machine learning for this investigation.

Main Methods:

• Utilized multiplexed imaging to capture detailed cellular and protein information.
• Applied machine learning algorithms to analyze the relationship between cell shape features and protein expression.
• Employed machine learning interpretability techniques to explore functional implications of shape features.

Main Results:

• Established a universal, bi-directional link between cell shape and protein expression across diverse cell types and disease states.
• Identified that this shape-protein link is amplified for cell state markers.
• Discovered a subpopulation of large, p53-positive tumor cells in two cancer types.
• Demonstrated that cell shape properties improve graph neural network-based disease state prediction.

Conclusions:

• Cell shape and protein expression are intricately coupled, offering insights into cellular function and disease.
• Shape features can guide hypotheses regarding protein function.
• This approach enhances understanding of tissue organization and state prediction in physiological contexts.