Background. Recent technological advances allow molecular profiles (e.g., gene expression) in single cells, while also retaining information of the spatial tissue. Spatial context of gene expression enables, among others, the study of cell migrations, cell-to-cell interactions, cell-type and gene expression spatial organization, and to link tissue structural architecture to functional organisation. The uptake of spatial omics technologies is leading to a rapid emergence of novel, ad hoc, methods; as more and more methods are developed, is it crucial to assess their performance. Since real datasets usually do not carry ground truth, simulations are typically used to evaluate and compare methods; nonetheless, spatial omics data currently lack general simulation frameworks.
Aim. Here, we aim to develop a novel statistical approach for simulating spatial omics data. Starting from a real dataset (used as anchor data), our framework will enable simulating accurate and realistic spatial omics profiles, for a wide range of technologies (i.e., spot- and single-cell level).
Impact. Our tool will allow generating synthetic datasets, which will be essential for comparing (present and future) methods' performance, and understanding i) the merits and limits of each tool, and ii) what methods work under what scenario.
Resources. The Robinson lab will provide the computational resources required to complete the project (i.e., access to multicore servers). Scripts will be run in R.
Deliverables. At the end of the project, a Bioconductor R package (for simulating spatial omics data) will be released, and a manuscript will i) appear as a pre-print and ii) be submitted to a peer-reviewed journal.
References. [1] Burgess, D. J. (2019) Spatial transcriptomics coming of age, Nature Reviews Genetics. doi:10.1038/s41576-019-0129-z [2] Marx, V (2021). Method of the Year: spatially resolved transcriptomics. Nat Methods. doi:10.1038/s41592-020-01033-y