Using deep learning models to infer spatial transcriptomics from H&E slides
In this project we aim to enhance and validate our current model path2space which infers spatial transcriptomics from breast cancer H&E slides. The goal is to expand and continue to validate our approach to be able to asses different types of cancer from H&E slides. This tool could help find spatial markers from large datasets that couldn’t be before.
We believe that the PLCO dataset is a good fit for our tool as it has comprehensive information regarding each of the patients . This information could further help us find biomarkers for different subgroups of the disease.
We have a few aims
1. Expand and validate our current path2space model in order to make it work on various types of cancer.
2. Assess different techniques that could help make a better model.
3. Find pan-cancer and cancer specific biomarkers that could help in turn find new disease pathways.
4. Find pan-cancer and cancer specific biomarkers for response to treatment or overall survival
5. Find new drug targets from the inferred spatial transcriptomics.
none at the moment