ARCADE disentangles cell abundance from continuous transcriptional state in spatial transcriptomics

Researchers have developed a new computational framework called ARCADE to better analyze spatial transcriptomics data, which is used to study both cell-type abundance and within-type transcriptional variation. The challenge lies in separating these signals when they are measured at the spot-level in mixed samples.

ARCADE tackles two primary sources of uncertainty in spatial analysis: unstable single-cell references and the entanglement between abundance and cell state. The framework begins by constructing a robust reference using hyperparameter optimization and consensus filtering. This reference is then utilized in a two-stage spatial inference process.

In the first stage, ARCADE estimates the composition of different cell types. In the second stage, it models the continuous intratype transcriptional programs around this compositional anchor.

Extensive testing using both simulated and semi-synthetic benchmarks has demonstrated that ARCADE significantly improves deconvolution accuracy compared to existing approaches. Moreover, the framework yields more stable state inference results.

When applied to matched human dorsolateral prefrontal cortex datasets that also include spatial proteogenomic readouts, ARCADE is able to discern localized cell enrichment from astrocyte and microglial activation programs. This observation highlights the potential of decoupled modeling to enhance the interpretability of spatial transcriptomic signals in complex tissues.