Spatial Transcriptomics through the lenses of statistical modeling and AI

In recent years technological advances have made it possible to quantify the mRNA expression of large numbers of genes while preserving the spatial context of tissues and cells.

These techniques, collectively known as spatial transcriptomics, are important because key biological processes depend on the physical proximity of cells and the spatial organization of tissues.

Furthermore, several diseases are characterized by abnormal spatial organization within tissues.Despite its early age, spatial transcriptomics is rapidly becoming a widely used tool, complementing single-cell RNA-seq as the tool of choice to study gene expression in complex tissues, e.g., in cancer research and neurobiology.

In addition to the gene expression measurements and the spatial localization of transcripts, available data include images collected from the samples that can be used to learn cell-level and tissue-level morphological features.

The main objective of this proposal is to combine transcriptomics, spatial structure, and morphology data to better inform key spatial transcriptomics analytical steps.Specifically, we will:- Enhance the comprehension of imaging-based spatial transcriptomics data by a characterization of the statistical properties of the data and a mechanistic modeling of the in situ transcriptional measurements.- Combine imaging and tabular data to better define cell types and states through the use of statistical models and artificial intelligence.- Develop an inferential framework to model the localization of transcripts within and across cells and of cells within and across samples.Overall, this proposal will combine machine learning and artificial intelligence approaches with rigorous statistical modeling of transcriptomics data in a spatial, sub-cellular context.

This will ultimately serve the biomedical community and provide a suite of tools that will help pave the way towards personalized medicine and computer-assisted pathology.