Machine-learning approaches to develop pre-clinical biomarkers using longitudinal clinical tracks from multiple healthcare organizations for the prediction of disease development and clinical outcomes
The best way to treat diseases is to prevent them or to perform early diagnosis that enables providing treatment before the diseases enter severe and non-reversible states. In addition - early diagnosis of diseases will enable the investigation of the biological mechanisms of the diseases at the early stage - and thus to develop novel therapeutic interventions. To enable such approach - biomarkers and models for early diagnoses are required. Current medical practice for many diseases is usually responsive for the development of disease instead of proactive efforts to prevent it. Minority of preventable diseases also lack more accurate biomarkers or predictive models that will discern individuals at increased risk.
The Hebrew University of Jerusalem, the Hadassah medical center, and Meuhedet HMO join efforts to create seamless medical records that represent the timelines of many diseases and the collection of genomic and clinical parameters from the patients' history in order to develop and enable early diagnosis of many medical conditions.
We would like to exploit the synergistic advantages of our local cohort and the advantages of UKBB cohort. Our cohorts are smaller but we have access to the patients, their samples and we can perform prospective follow-up easily. UKBB cohort is larger and this enables powerful application of machine learning predictive models. These models can focus our research to the more productive projects. In addition, UKBB database include genomic information for most patients. Identification of strong association and predictive models will enable us to focus on the more promising predictive models, to apply them to different population along with prospective follow-up. On the other hand - we will be able to go in the other directions and apply insights gleaned from basic and clinical research into the larger cohorts of UKBB in order to apply to a different population and validate on large scale. All these model will be published in peer reviews journals and will be available for application to other populations and countries.