While pure sequencing-based methods allow for the identification of prognostic markers that might drive disease progression, recent spatial omics approaches add the additional context of spatial organization of tissue, cell location and molecular expression. This enables the stratification of patients by new spatial markers, for example how much immune cells infiltrate into tumor tissue and relating these findings back to molecular expression. Our group uses the GeoMx technology to manually select regions of interest in tumor tissues, each region containing a few hundred cells, for which we collect bulk count data of proteins and/or transcripts. Using deep learning from images to upfront identify cellular neighbourhoods governing patient outcome would objectively inform selection of regions of interest for detailed spatio-molecular profiling using GeoMx. In this project you will work with multiplex immunofluoresence images from a large cohort of ovarian cancer patients. You will be using published neural network methods, like Naronet or Space-GM, to identify tissue niches which can predict clinical outcomes.
Contact: anna.sandstrom_gerdtsson@immun.lth.se
