Predicting adverse outcomes from genomic data using machine learning models and localization.
Principal Investigator:
Mr Kristijan Vukovic
Approved Research ID:
59571
Approval date:
June 3rd 2020
Lay summary
The aim of our project is building predictive models for various human diseases and conditions. To achieve this, we'll develop machine learning methods trained on genotyping data available within the UK Biobank. There have been many studies suggesting strong genetic background to various phenotypic traits. Much of this genetic background is likely captured by genotyping assays, which screen the most interesting and diverse regions of the human genome. In this 24-months project, we'll try to utilize the vast genotyping data available within the UK Biobank database in order to build predictive models of the associated traits and conditions. Our primary focus will be predicting organ-level adverse outcomes, but we will attempt to predict various other responses as well. We expect to obtain a wide range of machine learning models and we hope that some of them will have a high enough predictive performance as to make them useful for potential application in th erisk-assessment of various diseases. This type of method could improve public health outcomes by assessing the risk of the individual towards a certain disease or conditions.