Statistical and Computational Methods for Cell Type Deconvolution in Spatial Transcriptomics

Explore statistical and computational approaches for cell type deconvolution in spatial transcriptomics through this 40-minute conference talk by Ying Ma at the Computational Genomics Summer Institute. Learn about cutting-edge methods that identify and quantify different cell types within spatially resolved transcriptomic data, addressing one of the key challenges in spatial biology. Discover how spatially informed deconvolution techniques can improve upon traditional methods by incorporating spatial context and neighborhood information. Examine robust decomposition strategies for handling cell type mixtures in complex tissue environments and understand the practical considerations for implementing these methods in real-world datasets. Gain insights into benchmarking approaches that evaluate the performance of different deconvolution algorithms and learn about fine-grained cell type mapping techniques that can resolve cellular heterogeneity at unprecedented spatial resolution. The presentation draws from recent advances in the field, including developments in Bayesian approaches, machine learning methods, and comprehensive evaluation frameworks that guide method selection for specific experimental contexts.