Inferring causal trajectories from spatial transcriptomics using CASCAT.

Area of Science:

• Computational biology
• Genomics
• Systems biology

Background:

• Spatial trajectory inference is crucial for understanding cell differentiation and tissue dynamics.
• Current methods often rely on spatial proximity and overlook the Markovian property of cell state transitions.
• Challenges exist in inferring unique trajectories from high-dimensional, nonlinear data due to Markov equivalence.

Purpose of the Study:

• To introduce CASCAT, a tree-shaped structural causal model integrating the Markovian property for unique cell differentiation trajectory inference.
• To address limitations of existing methods in handling complex biological data.

Main Methods:

• Developed CASCAT, a novel tree-shaped structural causal model.
• Integrated the Markovian property into the causal inference framework.
• Applied CASCAT to analyze simulated and real single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics datasets.

Main Results:

• CASCAT outperformed six state-of-the-art scRNA-seq methods and three leading spatial trajectory inference methods.
• Accurately identified cell maturation trajectories and revealed the Wnt signaling pathway in the mouse inner olfactory bulb.
• Improved drug response prediction precision by 6.8% in oral squamous cell carcinoma compared to RNA velocity methods.

Conclusions:

• CASCAT provides a robust framework for inferring unique cell differentiation trajectories.
• Demonstrates significant improvements over existing methods in both accuracy and biological insight.
• Offers potential for advancing computer-assisted drug discovery and understanding complex biological systems.