A Guide for a Cardiovascular Genomics Biorepository: the CATHGEN Experience

Abstract

The CATHeterization GENetics (CATHGEN) biorepository was assembled in four phases. First, project start-up began in 2000. Second, between 2001 and 2010, we collected clinical data and biological samples from 9334 individuals undergoing cardiac catheterization. Samples were matched at the individual level to clinical data collected at the time of catheterization and stored in the Duke Databank for Cardiovascular Diseases (DDCD). Clinical data included the following: subject demographics (birth date, race, gender, etc.); cardiometabolic history including symptoms; coronary anatomy and cardiac function at catheterization; and fasting chemistry data. Third, as part of the DDCD regular follow-up protocol, yearly evaluations included interim information: vital status (verified via National Death Index search and supplemented by Social Security Death Index search), myocardial infarction (MI), stroke, rehospitalization, coronary revascularization procedures, medication use, and lifestyle habits including smoking. Fourth, samples were used to generate molecular data. CATHGEN offers the opportunity to discover biomarkers and explore mechanisms of cardiovascular disease.

Keywords: Air pollution; Biomarkers; Biorepository; Cardiometabolic disease; Cardiovascular disease; Genetics; Genomics; Geocoding; Metabolomics.

Figure 1. CATHGEN Ascertainment Timeline

DNA, plasma and PaxGene RNA tubes were collected on 9334 samples between January 2001 and December 2010. Timeline in counts per month are plotted with cumulative ascertainment also shown for total samples and distinct individuals, as some individuals were sampled more than once.

Figure 2. Age and gender of Sampled CATHGEN Individuals with Clinical Information

Aged of sampled individuals are shown by decades. The number of men and women in each category are shown by blue and green bars, respectively. The mean age for men was 62.0 (±12.5 SD) y and women was 60.8 (±11.7) y.

Figure 3. Geographic Distribution of the Residences of North Carolina CATHGEN Participants

Residences were assigned geocoded locations by the Duke School of the Environment using the declared residence of CATHGEN participants on the day of enrollment.

Figure 4. Venn Diagram of Subject Data and Overlap for Several Data Ontogenies

Of the 9334 subjects, CATHGEN contains air quality data on 8021; targeted metabolomics data on 3899; genome-wise SNP chip genotypes (GWAS) on 3649; genome-wide gene expression data on 1284; exercise stress test data on 2835; microRNA data on 705; and genome-wide DNA methylation data on 43 individuals. We also have lipoprotein profiling data on 8000 individuals (not shown).

Figure 5. Challenge and Approaches to Systems Modeling in Large Omics Datasets

In the upper panel is depicted the classical understanding of the hierarchy of regulatory control, wherein DNA leads to RNA to protein, to metabolite and then to phenotype. As shown in the middle panel and as explained in the text, in the case of CATHGEN, if attempting to pick out specific molecular regulatory pathways one is left with a untenable problem of trying to select the most likely regulatory events from a possible 10¹³ tests. Shown at each stage is the number of possible molecular probes (e.g., a typical gene expression array selects for approximately 20,000 gene expression targets). One approach is to use a phased approach (“Inverse QTL Modeling”); one works backward from phenotype toward gene, selecting significant targets at each stage, thus reducing the number of overall comparisons in the analysis. However, as shown in the lowest panel, this approach ignores the real hierarchy of regulatory control which is a not a linear process (e.g., metabolite produced by gene expression can have no role in the pathway from gene expression to phenotype; can mediate the effect of gene expression; or can feed back on gene expression — in a positive or negative fashion — and moderate the effect of gene expression on phenotype.