Multi-Omics Investigation into Acute Myocardial Infarction: An Integrative Method Revealing Interconnections amongst the Metabolome, Lipidome, Glycome, and Metallome

Abstract

Acute myocardial infarction (AMI) is a leading cause of mortality and morbidity worldwide. This work aims to investigate the translational potential of a multi-omics study (comprising metabolomics, lipidomics, glycomics, and metallomics) in revealing biomechanistic insights into AMI. Following the N-glycomics and metallomics studies performed by our group previously, untargeted metabolomic and lipidomic profiles were generated and analysed in this work via the use of a simultaneous metabolite/lipid extraction and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis workflow. The workflow was applied to blood plasma samples from AMI cases (n = 101) and age-matched healthy controls (n = 66). The annotated metabolomic (number of features, n = 27) and lipidomic (n = 48) profiles, along with the glycomic (n = 37) and metallomic (n = 30) profiles of the same set of AMI and healthy samples were integrated and analysed. The integration method used here works by identifying a linear combination of maximally correlated features across the four omics datasets, via utilising both block-partial least squares-discriminant analysis (block-PLS-DA) based on sparse generalised canonical correlation analysis. Based on the multi-omics mapping of biomolecular interconnections, several postulations were derived. These include the potential roles of glycerophospholipids in N-glycan-modulated immunoregulatory effects, as well as the augmentation of the importance of Ca-ATPases in cardiovascular conditions, while also suggesting contributions of phosphatidylethanolamine in their functions. Moreover, it was shown that combining the four omics datasets synergistically enhanced the classifier performance in discriminating between AMI and healthy subjects. Fresh and intriguing insights into AMI, otherwise undetected via single-omics analysis, were revealed in this multi-omics study. Taken together, we provide evidence that a multi-omics strategy may synergistically reinforce and enhance our understanding of diseases.

Keywords: acute myocardial infarction; glycomics; lipidomics; metabolomics; metallomics; multi-omics.

Figure 1

PLS-DA score plots for the visualisation of clustering of AMI versus healthy samples based on (A) lipidomics and (B) metabolomics analysis, and the corresponding model performance measures across the number of components generated from (C) metabolomics and (D) lipidomics analysis.

Figure 2

Distribution of lipid classes of annotated significant lipidomic features.

Figure 3

Visualisation of correlations amongst omics datasets via (A) sample scatterplot displaying the first component in each omics block (upper diagonal) and Pearson correlation between each component (lower diagonal), (B) correlation circle plot representing feature contributions from each omics block, and (C) circos plot showing the correlations (r > 0.33) between omics features as indicated by the red (positive correlation) and green (negative correlation) links.

Figure 4

Relevance network plots of significant omics features from the discrimination of AMI vs. healthy patients, displaying (A) key cluster amongst all four omics with strong correlations between features (r ≥ 0.67; p-value < 0.05), and (B) tri-omics (glycomics + metallomics + lipidomics) correlations (r ≥ 0.5; p-value < 0.05). The color key indicates the correlation coefficient values annotated by the connection lines between variables. Red colored connection lines denote positive correlations, while green colored connection lines denote negative correlations between variables. The intensity of the colors is scaled according to the magnitude of the correlation coefficient values.

Figure 5

Relevance networks of significant omics features across various bi-omics combinations (all r ≥ 0.4; p-value < 0.05). The color key indicates the correlation coefficient values annotated by the connection lines between variables. Red colored connection lines denote positive correlations, while green colored connection lines denote negative correlations between variables. The intensity of the colors is scaled according to the magnitude of the correlation coefficient values.

Figure 6

Classification performance of glycomics, metallomics, metabolomics and lipidomics datasets as shown from (A) block-PLS-DA analysis and comparison of individual blocks with a consensus multi-omics model, and (B) hierarchical cluster analysis of the multi-omics dataset.