Deep-learning data fusion for patient screening and clustering from genomics liquid biopsies and its application for presymptomatic cancer detection and stratified medicine

My PhD research relates to the design of an Artificial Intelligence (AI) framework for learning complex relations between the DNA profile of a liquid biopsy (i.e., DNA extracted from a blood sample) and the diagnosis, prognosis, or relapse of a tumor. This AI framework will be able to continuously integrate information from DNA profiles from tens of thousands of liquid biopsies. In particular, it will allow building models for the early diagnosis of cancer, by discriminating between healthy patients and patients with genomic DNA imbalances. Each model will have the ability to learn new predictive tasks as new types of disease status get revealed, as well as the ability to incorporate data from follow-up liquid biopsies. The inference process will handle both new patients and new data available for a same patient. Also, the latent representation learned by the AI platform will allow interpreting genomic profiles, clustering patients, and identifying common disease mechanisms. This, along with clinical expertise will be a powerful approach for modeling the evolution of diseases inside a subgroup of patients.