By Gene Yang ’19
Congenital heart disease is the most common birth defect in the world. This broad group of genetic conditions affect the heart’s structure and function in different ways, with symptoms ranging from harmless to fatal. However, very little is known about how a human’s genetic makeup affects cardiovascular development. A recent study, which constructed a model that predicts correlations between genetics and heart structure, addresses this discrepancy
Although recent advances in cardiac imaging techniques are promising, there currently exists no automated process for diagnosing heart disease. Thus, every cardiac imaging session involves a cardiologist who manually checks the imaging results for possible physical abnormalities, such as hypertrophy, thickening of the heart muscle, or bicuspid valves, when a heart valve possesses two leaflets instead of three. This process is not only time consuming, but also fails to elucidate the relationships between genetics and heart structure due to the limited amount of variables we can manually observe.
Thus, the study employed a high-spatial resolution software that converted the cardiac MRI images of 1,124 healthy volunteers into 3D cardiac models. Using this data, the researchers mapped associations between genotypes and cardiac phenotypes. This was done with mass univariate regression, a statistical test that utilizes a general linear model to compute the regression coefficient associated with a variable of interest for each vertex of the 3D image. Because analyzing each point as an independent variable produces more erogenous results, a statistical problem known as the multiplicity, a method called threshold-free cluster enhancement (TFCE) was then used to “cluster” spatially similar vertices together. Notably, researchers found significant correlation between cardiac wall thickness and four single nucleotide polymorphisms (SNPs), with regression coefficients ranging from -0.11 to 0.1.
This novel approach to cardiac genotype-phenotype mapping, although in its early stages of development, offers new insight into the ways our genes shape the development of our hearts. With future advances in machine learning and imaging techniques, finding the answer to this longstanding biological question seems promising.
- C. Biffi, et al., Three-dimensional cardiovascular imaging-genetics: a mass univariate framework. Bioinformatics 33, 90-96 (2017). doi: 10.1093/bioinformatics/btx552.
- Image retrieved from: http://www.greensbororadiology.com/portals/0/Images/GR%20photos/cardiac-screening.jpg