Statistical Bioinformatics Webinar: Miller -- Reference-free Cell-type Deconvolution of Multi-Cellular Pixel-Resolution Spatially-Resolved Transcriptomics Data

Presented by Dr.  Brendan Miller, Johns Hopkins University 

Recent technological advancements have enabled spatially resolved transcriptomic (ST)
profiling but at multi-cellular pixel resolution, thereby hindering the identification
of cell-type spatial co-localization patterns.  Supervised deconvolution approaches have
recently been developed to predict the proportion of cell-types within ST multi-cellular
pixels but these approaches rely on the availability of a suitable single-cell
reference, which may present limitations if such a reference does not exist.  To address
this challenge, we developed STdeconvolve as an unsupervised approach that builds upon
latent Dirichlet allocation to deconvolve underlying cell-types comprising such ST
datasets.  We show that STdeconvolve effectively recovers the putative transcriptomic
profiles of cell-types and their proportional representation within ST multi-cellular
pixels without reliance on external single-cell transcriptomics references.  We find
that STdeconvolve provides competitive performance to existing reference-based methods
when suitable single-cell references are available, as well as potentially superior
performance when suitable single-cell references are not available.