Spatial transcriptomics techniques and spatially annotated scRNA-seqdata setprovides unprecedented opportunities to analyze cell-to-cell communication (CCC). However, integrating the spatial information and complex biochemical processes required for CCC reconstruction remains a major challenge. Recently the journalNature Methods "Published.An optimal transport method for dealing with complex molecular interactions and spatial constraints: the COMMOT, inferring CCC in spatial transcriptomics.
What is COMMOT?
COMMOT (Communications Analysis for Optimal Transmission)CCC was inferred by simultaneously considering a large number of ligand-receptor pairs from either spatial transcriptomics data or spatially annotated scRNA-seq data equipped to infer CCC based on pairwise spatialimagingSpatial distances between cells estimated from the data; summarizing and comparing directions of spatial signals; using treemouldIdentify downstream effects of CCC on gene expression; and provide visualization tools for various analyses.
COMMOT Overview
COMMOT has three important features: first, the use of non-probabilistic mass distributions to control the margins of the transportation scheme to maintain comparability between species; second, the imposition of spatial distance constraints on the CCCs to avoid connecting spatially distant segments; and, finally, the transfer of multi-species distributions (ligands) to multi-species distributions (receptors) to account for multi-species interactions.
COMMOT performance testing
The developers applied COMMOT to simulated data and eight spatial datasets acquired using five different techniques to show its effectiveness and stability in identifying spatial CCCs in data with varying spatial resolution and gene coverage:By integrating computerized spatial transcriptomics data obtained from scRNA-seq and spatial staining data, Visium, Slide-seq, STARmap, MERFISH, and seqFISH+ spatial transcriptomics, COMMOT can consistently capture CCC activities known from the literature。In human skin, COMMOT showed that higher WNT signaling increased the expression of several genesThis result was confirmed by immunofluorescence staining.
The role of CCC in human skin development
COMMOT was compared to three methods for inferring CCC: CellChat, Giotto, and CellPhoneDB v3.By examining the correlation between inferred CCC and known downstream gene activity, the development team found that for most data sets, theCOMMOT has a stronger correlation than the three approaches, in some cases, the correlation between COMMOT and CellPhoneDB v3 was comparable. This assessment could be further improved if more complete knowledge of gene regulation becomes available.
Cluster-level correlations between inferred signaling and known downstream gene activity (compared to three methods for inferring CCC)
With the predictable availability of time series of spatial transcriptomics data, CCC dynamics can be elucidated, e.g., by extending collective optimal transport into dynamic optimal transport formulations.The PDE model of CCC can be extended to further incorporate intracellular gene regulatory networks. While traditional optimal transport is powerful in integrating pairs of datasets and multi-marginal optimal transport integrating multiple datasets, collective optimal transport can efficiently control coupling and deal with competing species, which is useful for a wide range of problems beyond CCC inference.
Code availability
COMMOT is available at GitHub: /zcang/COMMOT
bibliography
Cang, Z., Zhao, Y., Almet, . et al. Screening cell–cell communication in spatial transcriptomics via collective optimal transport. Nat Methods 20, 218–228 (2023). /10.1038/s41592-022-01728-4