Establishment of a federated database
RHAPSODY assembled a federated database with fully harmonized data (CDISC SDTM format) from 6 prediabetes cohorts, 4 T2D progression cohorts, 1 gastric bypass cohort and 1 clinical trial cohort, from 5 different EU countries and totalling > 68,000 individuals with their clinical data. We generated new C-peptide measurements, new lipidomics, peptidomics and metabolomics data and obtained genetics data for 3 cohorts. Data deposited in the database follow the FAIR (findable, accessible, interoperable, reusable) principles. In addition to several essential analytical tools, a new open access software, dsSwissKnife, has been written to enable cross-cohort statistical analysis. Workshops were organized for hands on learning to use the available tools. The federated database was actively exploited by all WPs and also used to discuss biomarker validation with the EMA and MHRC.
Multi-omics analysis for biomarkers selection
Plasma lipidomics data were obtained from 3 T2D progression cohorts (2,775 samples); quantitative plasma peptide and protein data from 1200 samples; polar metabolite data from > 5,000 samples. Polar metabolites and plasma lipidomics data were also generated from 268 plasma samples from pancreas surgery patients (partial pancreatectomy patients) for whom we also obtained clinical information and islet transcriptomic profiles. Plasma lipidomics and islet, liver, fat and muscle transcriptomic data were generated from mouse models of prediabetes. All data are deposited either in the federated database for sensitive clinical data, or in a central database for non-sensitive data.
Biomarker prioritization and validation
RHAPSODY established a biomarker prioritization matrix and a web-based prioritization tool enabling (i) to systematically evaluate biomarker candidates with relation to relevant external data sources and (ii) to place candidate biomarkers in the context of all RHAPSODY data generated and visualise results across multiple experiments. Together with the federated database, these resources provided a platform for biomarker discovery, key outcomes of which include:
• The use of the federated database to model T2D progression using genetics and omics data in combination with the clinical measurements allowed (i) refining the previously characterized 5 patient subgroups and (ii) identifying the underlying molecular mechanisms related to islets, liver and adipose tissue metabolism, which provide novel insights into the diverse aetiological processes.
• Paired plasma and islets samples were obtained from diabetic, glucose intolerant and non-diabetic patients undergoing partial pancreatectomy for islet transcriptome and proteome analysis and for plasma lipidomic analysis. Integrated data analysis identified islet gene co-expression modules that correlate with T2D progression and which provide a mechanistic model of beta-cell failure in T2D.
• Human islet transcriptomic, proteomic and lipidomic data analysis identified (i) islet cell mechanisms that lead to amyloid plaque deposition (T2D islets marker) and (ii) mechanisms by which gluco-lipotoxic conditions induce irreversible beta-cell dysfunction.
• Analysis of plasma lipidomics and islets and liver transcriptomic data from mouse models of prediabetes identified a link between plasma triglycerides, the liver ß-oxidation pathway and key genes controlling insulin secretion by beta-cells. Similar correlations were found using human lipidomic and islet transcriptomic data. This study also identified a novel regulator of insulin secretion. Thus, circulating triglycerides are part of a liver-to-beta-cell axis and are biomarkers of beta-cell function.
• Newly developed unsupervised and supervised multiblock analysis bioinformatic tools were used to visualize and analyze how changes in islet-liver-fat-muscle gene module interaction networks underlie multi-organ deregulation in T2D progression. Spt2 was one of the specific genes studied, revealing a link between ceramide production, bile acids, Fgf15 action and glucose homeostasis.
Modelling economic and public health impact of disease modification
RHAPSODY developed interactions with both the EMA and MHRC to seek advice about the use of RHAPSODY biomarkers in (i) predicting T2D development and progression and (ii) modifying disease drug treatments. Health economists performed cost-effectiveness assessments to evaluate the potential interest to EU and US health care systems of using RHAPSODY biomarkers to stratify or target preventive and therapeutic treatments of T2D.
