PULSAR: a Foundation Model for Multi-scale and Multi-cellular Biology

Area of Science:

• Computational biology
• Systems biology
• Precision medicine

Background:

• Biological systems involve complex interactions across multiple physical scales, from molecules to tissues.
• Existing computational models often analyze biological scales in isolation, limiting comprehensive understanding.
• Integrating multi-scale biological data is crucial for advancing health and disease research.

Purpose of the Study:

• To introduce PULSAR (Patient Understanding Leveraging Single-cell universal Representation), a multi-scale foundation model architecture.
• To enable seamless information flow from genes to cells to multicellular systems.
• To apply PULSAR to the human peripheral immune system for disease analysis and prediction.

Main Methods:

• Developed a multi-scale and multicellular foundation model architecture named PULSAR.
• Enabled explicit information flow across biological scales: genes, cells, and multicellular systems.
• Applied the model to the human peripheral immune system dataset.

Main Results:

• PULSAR extracts a unified donor representation from peripheral immune system data.
• The model achieves rapid disease classification, biomarker prediction, and clinical event forecasting (e.g., Rheumatoid arthritis onset).
• PULSAR simulates cytokine perturbation responses and identifies key disease-driving cell types.

Conclusions:

• PULSAR offers a novel computational approach to bridge molecular biology and clinical phenotypes.
• The model opens new avenues for precision medicine by enabling multi-scale biological reasoning.
• PULSAR facilitates a deeper understanding of immune system function in health and disease.