scverse Conference 2024

The last decade has been the witness of a meteoric development of multiplexed imaging (MI) technologies, allowing in theory the in-depth spatial profiling of biological tissues. Together with single-cell genomic assays, MI methods have thus the potential to shed light on poorly understood biological systems, ranging from developmental pathways to diseases pathogenesis. However, unlike single-cell genomic, the potential of MI methods is still constrained as it has not resulted in significant biological discoveries.

Here we show that the potential of MI can be unlocked by using tailored experimental sampling and data analysis approaches that embody the intrinsically spatial nature of MI. First, by coupling in-depth sampling simulation with a large scale spatial transcriptomic atlas, the rules governing MI experimental design were elucidated [1], and more efficient sampling strategies were established. Second, inspired by spatial ecology, we developed statistical models to analyze cell count derived from MI data, resulting in a drastic increase of statistical power for differential abundance analysis, as illustrated by the reanalysis of MI COVID-19 data [2]. Finally, we developed a new approach to model the distribution of cells through space in a quantitative manner, thus providing an interpretable feature extraction method to describe the shape of a group of cells. Combining it with the automated analysis and annotation of several MI RNA datasets, we identify a unique and recurring spatial pattern of immune infiltrating cells in auto-immune diseases, consisting in the aggregation of naive CD4-T cells into dense patches near TLS while activated CD8-T cells are homogeneously spread through the tissue.
Altogether, we provide a theoretical and practical basis for the experimental design and analysis of MI experiments, unleashing the potential of MI as a powerful technology for patient diagnosis and biological discovery, complementary to single-cell genomic.