Type 2 diabetes patients are a diverse group, and refining how they are classified would help individualise their treatment regimens. The course of the disease and the effectiveness of different medicines vary from one patient to another. It is also not clear why some at-risk people (such as those with obesity) develop the condition while others do not. Finding biomarkers would help in both identifying who is at risk of developing the disease, and determining individual patients’ likelihood of responding well to a particular treatment.
Main outcomes of DIRECT
The DIRECT consortium carried out deep phenotyping of patient groups from around Europe, in order to study extreme phenotypes of patients with very rapid or very slow glycaemic deterioration (either from prediabetes to diabetes, or through diabetes). They also sought to understand the extent to which certain therapeutic interventions result in improvement in glycaemia. They were able to perform clustering of type 2 diabetes-based and baseline characteristics, and explore the molecular signature associated with diabetes progression. It allowed them to identify processes that caused or contributed to the development of T2D and how they are linked to how the disease progresses.
Tests, clinical trials, omics
The project developed tests to predict who will get diabetes, whose condition will deteriorate rapidly after diagnosis, and who will respond well or badly to certain drugs. They also used machine learning to create prediction models for non-alcoholic fatty liver disease based on clinical data and omics data. They carried out analysis of plasma metabolite profiling at the beginning and again at 18 months, which gave them an objective measure of diet, as well as the link between glycaemic deterioration, cardio-metabolic health, pre-diabetics and T2D.
These biomarkers were tested in prospective clinical trials, paving the way for their use as new diagnostic tests as well as in the creation of personalised therapies. They enrolled approximately 2,300 pre-diabetics and 850 early T2D patients for extensive baseline and longitudinal, physiological, imaging and molecular phenotyping, and followed up at 48 and 36 months, respectively.
DIRECT established a multi-site trial infrastructure with a web-based accessible interface to allow for recruitment centres to upload and manage clinical and case report form data for all recruited patients.
Large-scale analysis of whole blood transcriptome revealed a large number of associations between transcriptomic modules and measures of insulin sensitivity and glucose tolerance, pointing to a major overlap of immune and metabolic processes. In a meta-analysis in genome-wide association studies that included approximately 5,500 subjects of European ancestry from six different cohorts treated with sulphonylureas, a gene variant with glycaemic response to sulphonylureas at a genome-wide scale was identified. They also identified 25 omics features from transcriptomics, metabolomics and proteomics data linked to early diabetes remission after obesity surgery.
Liver fat predictor
DIRECT developed liver fat prediction models (predictliverfat.org) and identified biological features that appear to affect liver fat accumulation. Non-alcoholic fatty liver disease (NAFLD) is highly prevalent and causes serious health complications in T2D. Early diagnosis is important, as this can help prevent irreversible damage to the liver and, ultimately, hepatocellular carcinomas. Using the baseline data from 1,514 participants, they expanded the etiological understanding and developed a diagnostic tool for NAFLD using machine learning. Multi-omic (genetic, transcriptomic, proteomic, and metabolomic) and clinical (liver enzymes and other serological biomarkers, anthropometry, and measures of beta-cell function, insulin sensitivity, and lifestyle) data were used as input.
Sulphonylurea and GLP-1R Agonist response
In large-scale analyses of trial and observational data, DIRECT identified novel mechanisms for why patients respond differently to two commonly used diabetes drugs: sulphonylulreas and GLP-1R Agonists. They identified genetic variants that alter response to these drugs – which provides insight into how these drugs work or are transported in the body, and importantly, have the potential to be used for targeted therapy in clinical management of diabetes in the near future.
Central database and biobank
During the project, all data was stored in a single secure server to enable high-performance analytics. The consortium built a central biobank to enable future biomarker discovery and replication for use in other IMI projects. The database with more than 40 terabytes of data is located in Denmark, while the biobank consisting of 300,000 samples is stored in the UK.
The work carried out under the DIRECT project boosted the industry’s understanding of the underlying causes of T2D, and is helping it to develop tailored treatments that can be targeted to the right patients. The work carried out in DIRECT complements the efforts of other IMI diabetes projects IMIDIA and SUMMIT.