Harnessing machine learning for a multi-omics approach to cardiovascular disease

Despite major advances in our understanding and treatment of cardiovascular disease (CVD), there are likely targets beyond the traditional lipid measurements that can better inform on diagnosis and prognosis of CVD.

We have the technical expertise and infrastructure to perform multi-omics comparisons of proteins, lipid species and non-coding RNAs.

We want to build on this competitive advantage by leveraging machine learning techniques along with sparse estimation of correlation matrix that uses proximal gradient algorithm to compare biomarkers from different molecular entities.

We will apply a new computational pipeline combining machine learning with statistical modelling and optimization techniques for the analysis and association of available multi-omics data from four cohorts namely, two community-based cohorts with 15-25-years of longitudinal follow-up, a well characterised patient cohort with different subtypes of myocardial infarction and a large collection of human carotid endarterectomies.

We are anticipating that harnessing machine learning and multi-omics data can achieve better cardiometabolic phenotyping, distinguish between different aetiologies of myocardial injury and identify clinically relevant sub-phenotypes of atherosclerosis.