GwSPADE: reference-free deconvolution in spatial transcriptomics with gene weighting

Music City Center 

Most spatial transcriptomics technologies (e.g. 10x Visium) operate at the multicellular level, where each spatial location often contains a mixture of cells with heterogeneous cell types. Thus, effectively deconvolving cell-type compositions is critical for downstream analysis. Although reference-based deconvolution methods have been proposed, they depend on the availability of reference data, which may not always be accessible. Additionally, within a deconvolved cell type, cellular heterogeneity may still exist, requiring further deconvolution to uncover finer structures for a better understanding of this complexity. Here we present gwSPADE, a gene expression-weighted reference-free SPAtial DEconvolution method for spatial transcriptomics data. gwSPADE requires only the gene count matrix and employs appropriate weighting schemes within a topic model to accurately recover cell-type transcriptional profiles and their proportions at each spatial location, without relying on external single-cell references that may introduce batch effects. gwSPADE demonstrates scalability across various platforms and outperforms existing reference-free deconvolution methods such as STdeconvolve.

Deconvolution

Reference-free

Latent Dirichlet allocation model

Weighting scheme 

Biometrics Section