Predicting single-cell ploidy from digital pathology images in pan-sarcoma

Sarcomas are a diverse group of rare cancers that originate from mesenchymal tissues. The rarity and intrinsic genetic complexity of sarcoma have made it one of the most challenging tumours to diagnose and consequently, to treat. A key genetic characteristic of many sarcoma subtypes is aneuploidy. Aneuploidy is defined by losses and gains of chromosomes, and is a prognostic marker for a variety of sarcomas. Currently, tumour ploidy are most commonly estimated using copy number alteration calling software on whole-exome and whole-genome sequencing data. However, the unsatisfactory accuracy, economic costs, infrastructure requirements, lack of trained personnel in interpreting the results is still an impediment to global adoption of these techniques for routine use, particularly in less developed economies. The goal of this research is to develop a more accessible and more accurate approach for predicting tumour ploidy based on Hematoxylin and Eosin (H&E) stained whole slide images. The specific aims of the project are to 1) optimise a nucleus segmentation model for annotating sarcoma nuclei in H&E images, and predict tumour ploidy at single-nucleus level, 2) thoroughly validate image ploidy and systemically compare it with existing ploidy estimation methods, 3) evaluate the diagnostic and prognostic value of the image ploidy, and investigate its underlying biological mechanisms in pan-sarcoma.This work will be carried out at University College London and the Royal National Orthopaedic Hospital, under the supervision of Prof. Adrienne Flanagan and Dr. Nischalan Pillay. The host labs have collected and scanned whole slide images of more than 40 types of bone and soft tissue sarcomas from more than 8000 cases, of which 1,600 cases have paired WGS and transcriptome sequencing. The access to this unique dataset will give an unprecedented opportunity to study ploidy at single-nucleus level in pan-sarcoma.