An algorithmic perspective on deciphering cell–cell interactions with spatial omics data

TL;DR

This review explores how spatial omics data and algorithms help understand interactions between cells in tissues.

Contribution

The paper categorizes and reviews existing algorithms for analyzing cell-cell interactions using spatial omics data.

Findings

Spatial CCI methods can be grouped into supervised learning, statistical correlation, and optimization approaches.

These methods are used for tasks like refining CCI networks and analyzing signal propagation in tissues.

The paper suggests future directions to improve current CCI methods using spatial data advancements.

Abstract

The advent of technologies to measure molecule information from a tissue that retains spatial information paved the way for the development of cell–cell interaction (CCI) methods. Even though these spatial technologies are still in their relative infancy, the developed methods promise more accurate analysis of CCIs due to the inclusion of spatial data. In this review, we outline these methods and provide a high-level view of the algorithms they employ. Moreover, we investigate how they deal with the spatial nature of the data they use and what types of downstream analyses they execute. We show that spatial CCI methods can broadly be classified into supervised learning, statistical correlation, and optimization methods that are used for either refinement of CCI networks, spatial clustering, differential expression analysis, or analysis of signal propagation through a tissue. In the end, we highlight some avenues for the development of complementary CCI methods that exploit advances in spatial data or alleviate certain downsides of the current methods.

Figures