Researchers develop new system to classify endocrine hypertension

The researchers, Dr Parminder S. Reel and Dr Smarti Reel, led by Professor Emily Jefferson from Dundee’s School of Medicine and Director of the Health Informatics Centre, developed a machine learning based pipeline for distinguishing different hypertension subtypes using multi-omics data and identifying their specific profiles.

Data generated from analyses of different biological sources, such as blood and urine, from the same individual is referred to as multi-omics data.  Combining these complex biological data can help in identifying links and biomarkers to build elaborate markers of disease and physiology.

Arterial hypertension is one of the major chronic diseases causing high morbidity and mortality. Around the world, approximately 25% of the adult population is affected by hypertension, which represents a tremendous public health burden. Hypertension carries an increased risk for various cardiovascular and renal diseases, such as stroke, heart failure, and chronic kidney disease.

To treat patients with hypertension, it is crucial to know the disease subtype. Endocrine forms of hypertension of adrenal origin are different from the more prevalent primary hypertension.

Currently, stratifying different hypertension subtypes entails lengthy evaluation involving blood/urinary tests, imaging procedures and eventually invasive testing. The delay in diagnosis can sometimes take several years after the onset which exposes patients to an increased risk of renal and cardiovascular damage and a diminished quality of life.

The machine learning based innovative approach to stratification is an advancement toward the development of a diagnostic tool which can significantly increase testing throughput and accelerate the administration of appropriate treatment.

In the retrospective phase of the research, around 500 patients with endocrine and primary hypertension and 100 non-hypertensive controls were included from existing patient registries from specialist centres for adrenal disorder across Europe. The omics measurement centres then used advanced techniques to measure different omics features in the blood/urine samples of these patients. These omics datasets are then integrated and evaluated using a machine learning framework for identifying a signature that enables us to distinguish patients with different hypertension subtypes.

Prof Emily Jefferson said, “The ENS@T-HT study is currently capturing data from a wider population in a prospective manner to measure the most discriminating features of the new samples and perform independent validation.

“This would allow the classifier to become more robust and well-trained for a formal clinical deployment. We have also evaluated the validity and cost-effectiveness of using the omics-based protocol for stratifying hypertensive patients, to open the way for implementation of this approach in clinical practice.”

Dr Parminder S. Reel added, “Until now most of the hypertension stratification studies have focused on mono-omics analysis. This is the first-ever study to investigate multidimensional omics data from a large cohort of hypertension patients using machine learning”

Prof Maria-Christina Zennaro (INSERM), coordinator of ENS@T-HT project said, "Multi-omics integration is a logistically challenging task since biosamples are sourced from multiple recruitment sites and require multi-centre omics measurements. This can lead to fewer samples with all available omics for integration. The ENS@T-HT study, by obtaining a complete set of omics for ~84% of the total patients, provided a straightforward example that this challenge can be successfully addressed. A major strength of this study was to rely on unambiguous diagnosis according to guidelines by expert centres."

More details on this project are available here.