Data-Driven Insights into the Association Between Oxidative Stress and Calcium-Regulating Proteins in Cardiovascular Disease

Figures

• 

  Figure 1. Overview of the workflow. This workflow illustrates the process of extracting, transforming, loading, and analyzing the relevant data sources. The leftmost column represents the data sources: UniProt for the Ca2+-regulating proteins, MeSH for oxidative stress (OS) concepts and 8 categories of cardiovascular diseases (CVDs), PubMed (https://pubmed.ncbi.nlm.nih.gov/) for documents, and Reactome (https://reactome.org/) for relevant pathways. Two text cubes were assembled for collections of relevant documents studying OS and CVDs respectively. CaseOLAP scores were computed to quantify the relevance with respect to the CVD categories for every relevant protein. This information was integrated via the construction of a knowledge graph, along with documents, MeSH descriptors, and pathways.

• 

  Figure 2. Knowledge-graph data schema. This data schema represents the knowledge graph (KG)’s structure, its node, and edge types. There are 4 nodes (MeSH, document, protein, and pathway), as well as 3 edges (document–assigns–MeSH, document–mentions–protein, and pathway–contains–protein).

• 

  Figure 3. Visualization of document-level association and molecular-level interaction between oxidative stress (OS) and cardiovascular disease (CVD) categories. (A) The top panel is a Sankey diagram outlining the document-level OS-CVD association. The column height represents the total number of documents associated with each OS phase and CVD category; the chord thickness is defined by the proportion of documents shared between each OS-CVD pair. An interactive version of this plot is also available online (https://caseolap.github.io/IonChannel/rcodes/docs-intersection.html). The bottom panel shows the detailed document counts for the Sankey diagram above. (B) The top panel is a Sankey diagram representing the molecular-level interaction between OS and CVD categories that are specifically relevant to Ca2+-regulating proteins. The column height represents the total number of proteins related to each OS phase and CVD category, and the thickness of the chord represents the number of proteins shared among each OS phase and CVD category. The interactive version is available online (https://caseolap.github.io/IonChannel/rcodes/protein-intersection.html). The bottom panel provides the numbers that…

• 

  Figure 4. Analysis of top-scoring proteins and PCA results: (A) demonstrates the top-scoring proteins across all cardiovascular disease (CVD) categories, visualized as a stacked bar chart; and (B) illustrates the results of the protein’s principal component analysis (PCA) within the 8-dimensional vector space corresponding to the CVD categories. This is subsequently projected onto a 2-dimensional plane defined by the primary principal components (PC1 and PC2). Individual dots signify distinct proteins, while the arrow-headed vectors represent the eight CVD categories in relation to PC1 and PC2. This illustration helps in comparing the behavior of the CVD categories based on the scoring proteins, providing insights into potential groupings of proteins operating via analogous molecular mechanisms. Notably, cardiac conduction system disease (CCS) and arrhythmias, cardiac (ARR), emerge as distinct from the remaining CVDs. (C) indicates the contribution of each CVD within the PCA dimensions. Considering the first two dimensions (i.e., PC1 and PC2), the first dimension is the positive superposition of all diseases, and ARR and CCS dominate the second dimension. The color bar in (A–C) represents…

• 

  Figure 5. Clustering behaviors of CVD proteins. The cluster image demonstrates the hierarchical clustering of protein scores into a two-dimensional cluster plot. The cluster is formed based on the Euclidean distance metric in 8-dimensional protein vector space. The distance scores obtained are distributed in a cluster plot ranging from −0.8 to 1.0, as shown in the color legend. A darker intensity indicates closely clustered proteins. The results demonstrated two significant clusters. The cluster enclosed with a red square represents a group of proteins associated with arrhythmias, cardiac (ARR), and cardiac conduction system disease (CCS). The cluster enclosed with a yellow square represents a group of proteins related to cardiomyopathies and heart failure (CM), myocardial ischemia (IHD), and other heart diseases (OHDs). The bar plot on the right is the visualization of the CaseOLAP scores of proteins rearranged based on the clustering.

• 

  Figure 6. Significant OS molecules and pathways in an ARR-CCS cluster: This figure presents sunburst visualizations illustrating significant oxidative stress (OS) molecules (A) and pathways (B) linked to the proteins identified in the arrhythmias, cardiac (ARR), and cardiac conduction system disease (CCS) clusters. Quantitative assessments of these OS molecules and pathways were conducted by calculating their respective scores based on the CaseOLAP scores of proteins within each cardiovascular disease (CVD) category. For OS molecules, scores were determined using the formula inside the parenthesis (OS molecule score = mean (protein’s CaseOLAP score in each CVD category)). For pathways, the score was computed utilizing the formula inside the parenthesis (pathway score = mean (protein’s score in each CVD category ∗ (1 − p-value)). In both (A,B), the arc thickness symbolizes the cumulative score for each molecule and pathway across the eight CVD categories. The outermost ring of the sunburst represents individual OS molecules (A) and pathways (B), with the arc width reflecting the score magnitude. The successive inner rings are organized hierarchically, with (A) categorizing candidate…

• 

  Figure 7. Link prediction analysis implementing graph algorithm: The heatmap depicts the predictive association scores derived from link prediction algorithms between oxidative stress (OS) molecules and Ca2+-regulating proteins (A), as well as between Ca2+-regulating proteins and cardiovascular diseases (CVDs) (B). The color intensity within the heatmap corresponds to the strength of the predicted association, with darker shades indicating a higher likelihood of connection between entities. Notably, proteins such as cardiac troponin I, nitric oxide synthase, matrix metalloproteinase-9, and gap junction alpha-1 protein, which play a role in calcium regulation, are identified with a high degree of probability as being associated with both OS molecules and CVD, underscoring their potential relevance in the underlying molecular mechanisms.