DeST-OT: Alignment of Spatiotemporal Transcriptomics Data

Peter Halmos,Xinhao Liu,Julian Gold,Feng Chen,Li Ding,Benjamin J. Raphael

DOI: https://doi.org/10.1101/2024.03.05.583575

2024-03-10

Abstract:Spatially resolved transcriptomics (SRT) measures mRNA transcripts at thousands of locations within a tissue slice, revealing spatial variations in gene expression and distribution of cell types. In recent studies, SRT has been applied to tissue slices from multiple timepoints during the development of an organism. Alignment of this transcriptomics data can provide insights into the gene expression programs governing the growth and differentiation of cells over space and time. We introduce ( velopmental patio emporal ptimal ransport), a method to align SRT slices from pairs of developmental timepoints using the framework of optimal transport (OT). uses optimal transport to precisely model cellular growth, death, and differentiation processes that are not well-modeled by existing alignment methods. We demonstrate the advantage of on simulated slices. We further introduce two metrics to quantify the plausibility of a spatiotemporal alignment: a which quantifies the discrepancy between the inferred and the true cell type growth rates, and a which quantifies the distance traveled between ancestor and descendant cells. outperforms existing methods on these metrics in the alignment of spatiotemporal transcriptomics data from the development of axolotl brain.

Bioinformatics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to align spatio - temporal transcriptome data at multiple time points during development to reveal the laws of cell growth and differentiation in space and time. Specifically, the authors proposed a method named DeST - OT (Developmental SpatioTemporal Optimal Transport) for aligning spatially - resolved transcriptome (SRT) slices from different developmental time points. This method precisely models the processes of cell growth, death, and differentiation through the optimal transport (OT) framework, especially the semi - relaxed optimal transport, which are often not well - handled in existing alignment methods.

### Main problems
1. **Cell growth and differentiation**: During development, cells in tissues will experience growth, death, and differentiation, and these processes lead to complex interactions between SRT slices at different times. Traditional alignment methods usually assume that the number of cells remains unchanged and cannot handle these dynamic changes well.
2. **Spatial deformation**: Tissues may expand or contract during development, causing the shapes of slices at different time points to change. Existing alignment methods often rely on static shape matching and are difficult to adapt to this spatial deformation.
3. **Multi - scale interaction**: The interactions between cells are not only the relationships between individual cells, but also involve complex interactions between multiple cells. Existing methods can usually only capture lower - level interactions and ignore higher - level interactions.

### Solutions
DeST - OT solves the above problems in the following ways:
- **Semi - relaxed optimal transport**: It relaxes the strict constraints on marginal distributions in traditional optimal transport, allowing some positions in the slice at the first moment to contribute more mass (representing cell growth), while other positions contribute less mass (representing cell death). This enables DeST - OT to better simulate the dynamic changes of cells.
- **Multi - scale interaction modeling**: The objective function of DeST - OT includes three parts: the doublet term, the triplet term, and the quartet term. The doublet term compares the gene expressions at corresponding positions in two slices; the triplet term models the relationships between ancestral cells and multiple descendant cells; the quartet term compares the spatial and transcriptome distances between two pairs of positions. These terms jointly capture the low - order to high - order interactions between cells.
- **Growth and migration metrics**: To evaluate the rationality of the alignment results, the authors introduced two metric indicators: the growth distortion metric and the migration metric. The growth distortion metric measures the difference between the inferred cell growth and the actual change in cell - type proportions; the migration metric measures the distance that cells move after alignment. These two metrics help to verify the biological and physical rationality of the alignment results.

### Experimental verification
The authors verified the effectiveness of DeST - OT on simulated data and real - world data of axolotl brain development. The experimental results show that DeST - OT is superior to existing methods in aligning cell - type boundaries, capturing cell growth and death, and handling noise.

In conclusion, this paper successfully solves the key problems in the alignment of spatio - temporal transcriptome data by proposing the DeST - OT method, providing a new tool for studying the dynamic changes of cells during development.