Spatial omics provides unpresidential profiling of tumors which can be useful to predict e.g. patient survival and drug response. The Spatial omics methods are however costly and complex thus currently not applicable in the clinical setting. To combat this issue, we have developed a deep learning algorithm “Image2Count” that learns from spatial omics data to predict molecular marker expression from just low-plex immunofluorescence tissue staining. In this project you will apply our developed method on single cell spatial transcriptomics datasets (CosMx or Xenium) to further validate the performance of Image2Counts. You may also use predicted expression data to model patient outcomes.
Contact: anna.sandstrom_gerdtsson@immun.lth.se
