StPedf: Cell trajectory inference of spatial transcriptomics via spatial proximity embedding and spatial...

Area of Science:

• Genomics
• Computational Biology
• Developmental Biology

Background:

• Spatial transcriptomics reveals gene expression heterogeneity within tissues.
• Spatial trajectory inference reconstructs cellular developmental paths using gene and spatial data.
• Existing methods struggle with nonlinear gene-expression and spatial coordinate interactions.

Purpose of the Study:

• To introduce StPedf, a novel method for spatial trajectory inference.
• To accurately model complex nonlinear interactions between high-dimensional gene expression and spatial coordinates.
• To enhance the accuracy and interpretability of reconstructing cellular differentiation trajectories.

Main Methods:

• Utilized a neural network with a masking mechanism to capture complex nonlinear interactions.
• Incorporated spatial proximity information as a guiding cue, adaptively adjusting embeddings and weights.
• Employed optimal transport to derive intercellular transition matrices and reconstruct trajectories.

Main Results:

• StPedf demonstrated superior performance over existing methods on five simulated datasets.
• Successfully mapped distinct lineages in telencephalon regeneration and malignant lineages in tumors.
• Generated developmental spatial trajectories and pseudo-spatiotemporal maps for human DLPFC.

Conclusions:

• StPedf significantly enhances accuracy and interpretability in spatial trajectory inference.
• Provides critical technical support for revealing dynamic cellular fate transitions in tissue microenvironments.
• Advances the understanding of cellular spatial organization in development and disease.