Mortality Prediction through Machine Learning
Principal Investigator: Professor Karsten Borgwardt
Approved Research ID: 14762
Approval date: October 1st 2015
Our research will focus on the topic of mortality prediction: Can the death risk of a patient in the near/mid future be predicted from genotypic data and current health status data? The UKB is an excellent data source for examining this connection, as it provides genotypic data (`Genomics`), current physical & health-state data (`UK Assessment Centre data`), and information on the death of participants (`Death Registry`). We focus here on mortality as a phenotype. Of course, this implicitly includes all health conditions that have been recorded as cause of death in at least one patient in the UKB. Being able to predict mortality is at the heart of improving the prevention, diagnosis and treatment of a wide range of serious and life-threatening illnesses, the stated goal of the UK Biobank. We will employ machine learning algorithms, that is computer protocols that detect patterns in data, to find out if there are clear discrepancies between the genetic & health record data of persons that are going to die within a certain time interval and those who are going to survive. Full cohort for which genotypic data is or will be available.