About the project:
Intra-abdominal malignancies often result in intraperitoneal free cancer cells (IPCCs), which increase the chances of cancer spreading to distal organs, and serve as an important prognostic tool in cancers such as ovarian and gastric cancers. Curative treatments based on intraperitoneal chemotherapy often have effective outcomes, however, the time between IPCC sampling and detection is critical, currently taking weeks to reach a conclusion. The MinION (Oxford Nanopore Technologies) – the first handheld genetic sequencer – is capable of reading long stretches of DNA in a real-time, but has not yet reached the level of accuracy of older, slower technologies. Deep learning techniques mimic the learning process of the human brain in order to recognize patterns in digital representations. In our lab, we use deep learning to circumvent the limitations of Nanopore sequencing by learning the ‘signal’ rather the ‘sequence’ of the DNA.
The aim of our project is to establish a rapid real-time method for the detection of IPCCs during colorectal resection. We propose a solution that takes advantage of our close collaboration with the Surgery Division at Sourasky Medical Center, and is based on three components: (i) access to clinical samples (abdominal fluid and mouth swabs) collected during resection; (ii) rapid real-time DNA sequencing; and, (iii) deep learning algorithms for the discrimination between somatic versus non-cancerous DNA. We believe the time has come to merge DNA sequencing (as a ‘digital signature’) with deep learning in order to afford surgeons the opportunity to quickly identify IPCCs during surgical procedures, thereby allowing immediate treatment and decreasing the need for future intervention.