Multi-omics approaches for revealing the complexity of cardiovascular disease

Abstract

The development and progression of cardiovascular disease (CVD) can mainly be attributed to the narrowing of blood vessels caused by atherosclerosis and thrombosis, which induces organ damage that will result in end-organ dysfunction characterized by events such as myocardial infarction or stroke. It is also essential to consider other contributory factors to CVD, including cardiac remodelling caused by cardiomyopathies and co-morbidities with other diseases such as chronic kidney disease. Besides, there is a growing amount of evidence linking the gut microbiota to CVD through several metabolic pathways. Hence, it is of utmost importance to decipher the underlying molecular mechanisms associated with these disease states to elucidate the development and progression of CVD. A wide array of systems biology approaches incorporating multi-omics data have emerged as an invaluable tool in establishing alterations in specific cell types and identifying modifications in signalling events that promote disease development. Here, we review recent studies that apply multi-omics approaches to further understand the underlying causes of CVD and provide possible treatment strategies by identifying novel drug targets and biomarkers. We also discuss very recent advances in gut microbiota research with an emphasis on how diet and microbial composition can impact the development of CVD. Finally, we present various biological network analyses and other independent studies that have been employed for providing mechanistic explanation and developing treatment strategies for end-stage CVD, namely myocardial infarction and stroke.

Keywords: cardiovascular disease; genome-scale metabolic model; integrated networks; omics integration; systems biology.

Figure 1

Schematic pathway for systems biology studies that utilize different omics data (genomics, transcriptomics, proteomics and metabolomics) to (A) further elucidate the complex aetiology and systemic effects arising from different tissues (heart, liver and visceral adipose) that contribute towards the development of atherosclerosis, and (B) investigate endothelial cell responses to flow and determine how different fluid shear stress states can promote either an ‘atheroprotective’ or ‘atheroprone’ state.

Figure 2

(A) Several signalling pathways in platelets converge on intracellular calcium release. (B) A dynamic neural network is used to train platelet response to combinatorial agonist activation. An input layer consisting of agonist concentrations is introduced to the 2-layer network at each time point and processing layers integrate input values with feedback signals for prediction of [Ca2+]i at the next timepoint. (C) Pairwise agonist synergy scores reflect gains or losses in calcium response due to agonist cross talk and are calculated for both experimental and predicted time-course traces to assess efficiency of the neural network. (D) Multiscale model of platelet deposition and thrombus formation under flow—this requires simultaneous solution of (1) intracellular platelet state ([Ca2+]i) and release reactions (R) for ADP and TXA2 calculated by Neural Network, (2) the instantaneous velocity field over a complex evolving platelet boundary □(t) calculated by Lattice Boltzmann, (3) concentration fields of ADP and TXA2 calculated by Finite Element Method, and (4) all platelet positions and attachment/detachment by Lattice Kinetic Monte Carlo.

Figure 3

The integration of omics data and application of systems biology approaches towards providing mechanistic insights on heart failure. (A) Formation of a condition-specific co-expression network through integration of gene expression profiles with protein–protein interaction data and biological function annotations enable the identification of functional modules representative of different biological processes which are relevant towards the progression of heart failure. (B) A comparative analysis between identified functional modules can reveal dynamic variations in the modules between healthy and disease states which can be used to classify normal and disease samples, thus implying plausible molecular mechanisms that are involved in the progression of heart failure. (C) Construction of a cardiomyocyte metabolic model (iCardio) through integration with protein data and manual curation from metabolic tasks. Functional metabolic changes are identified from gene expression data using metabolic tasks. (D) The application of GPR rules in determining reaction weights based on gene expression data, thereby assigning the expression of one gene as governing the reaction. The selected genes can differ between different datasets while producing the same net result, as seen here with a statistically significant decrease in the conversion of arginine to nitric oxide.

Figure 4

(A) Overview of a systems biology approach that links singular biomarker candidates towards deriving functional dependencies among chronic kidney disease (CKD) and cardiovascular disease (CVD). A set of genes associated with CVD in patients with CKD is used to construct a network of interacting proteins using reference data on known protein interactions. The functional interplay between interacting proteins was estimated by linking properties reflected by gene ontology terms, gene expression data characterizing CKD and TF binding sites with this methodology enabling detection of highly connected subnetworks associated with CKD and CVD. (B) The application of a kidney-specific metabolic model in determining metabolic process that are upregulated and downregulated in diabetic kidney disease resulting in the uptake and secretion of metabolites which may have an effect on the development and progression of CVD. Construction of the model was achieved through integration of proteome data from the Human Protein Atlas, transcriptome data from the Gene Expression Omnibus and application of a model-building algorithm that utilized Recon1 as a template to select only genes that were relevant to kidney.

Figure 5

(A) The application of different fields of study (metagenomics, metabolomics and transcriptomics) in assessing how the impact of diet can alter the composition of the gut microbiota, plasma metabolite concentrations and expression of genes which protect against or promote biological processes related to cardiovascular disease. (B) Investigating the interrelationship between gut microbiota composition, host metabolic profile and main CAD phenotype. The OTU-level network indicates the abundance of each OTU in a CAG based on node size. Specific CAGs that are mainly composed of certain OTUs correlate with serum metabolite concentrations which in turn correlate with parameters representing CAD severity. Red connections indicate a positive correlation and blue connections represent negative correlations according to Spearman correlation test (FDR < 0.05). In the CAG column, the green boxes indicate CAGs that were highly enriched in the control group and the purple box represents a CAG that was increased in the severe CAD group. In the metabolomics column, the pink boxes represent CAD-negative metabotypes and the yellow box represents a CAD-positive metabotype.