Skip to main content
Premium Trial:

Request an Annual Quote

Cardio Diagnostics Publishes Validation Data for Molecular Coronary Heart Disease Test

NEW YORK – Building on previous studies conducted with data from the Framingham Heart Study, new research published in the Journal of the American Heart Association found that Cardio Diagnostics' machine learning-based genetic-epigenetic test can help diagnose coronary heart disease.

The firm, which went public last year, has previously published work showing that its methods can predict three-year risk for the onset of symptomatic CHD, but the new study validates use of its methods for detecting stable CHD. Alongside collaborators from Intermountain Healthcare and the University of Iowa Hospitals and Clinics, Cardo Diagnostics researchers utilized an improved machine learning approach and included data from two additional cohorts with 449 cases of heart disease and 2,067 controls to develop a better machine learning model for determining symptomatic coronary heart disease (CHD).

The researchers noted that stable CHD is the result of multiple factors, including both genetic contribution and environmental stressors or lifestyle factors, such as smoking or a sedentary lifestyle. "Many of these risk factors have their own unique DNA methylation signatures in peripheral whole blood, often at key loci being affected by interactions between local or distant genetic variation," they wrote.

In its previous work, the research team used data from 1,545 subjects and a machine learning method to generate a random forest classifier incorporating the methylation information from four cytosine-phospho-guanine sites, two single nucleotide polymorphisms, age, and sex to predict symptomatic CHD. The accuracy of the test was 78 percent, with sensitivity at 75 percent and specificity at 80 percent. 

This previous work had multiple limitations, including the relatively small number of patients with CHD in the cohort, the demographic makeup of the study participants, and the fact that the CHD assessments were only best estimates. In addition, the methylation information used in the analysis was acquired from genome-wide arrays, "which are time consuming to process, costly, and relatively inaccurate," the researchers said.

In the new study, the team included data from three independent cohorts — the Framingham Heart Study, the Intermountain Healthcare Heart Institute registry, and the Iowa CHD Repository — and translated the DNA methylation findings into methylation-sensitive digital PCR (MSdPCR) assays. The end product, called PrecisionCHD, comprises a panel of these PCR assays and genotyping assays, as well as the machine learning algorithm. 

"The primary goal of this study was to develop a clinically implementable integrated genetic–epigenetic test that consists of standard fluorescent genotyping assays for SNP assessment, stand‐alone MSdPCR assays for methylation assessment, and a machine learning prediction model for CHD status prediction," the researchers wrote.

The machine learning model used a total of 10,484 methylation markers and 67,749 SNP markers for data mining. The Framingham cohort was divided into a training set consisting of 183 CHD patients and 1,400 controls and a test set with 61 CHD patients and 467 controls. 

When developing the algorithm, the researchers utilized digital methylation assessments of DNA samples already profiled with arrays "to renormalize the distribution of the array data at select loci." The methylation probes from the top performing sets were used to construct the MSdPCR tests and the markers were then used to determine methylation status of the entire Intermountain cohort, they said. The researchers developed an imputation model between array-based and MSdPCR-based methylation values and then conducted a feature selection step on the Framingham training set. 

The best performing feature set consisted of six methylation and 10 SNP markers, which were translated into standard MSdPCR assays and hydrolysable fluorescent primer probe genotyping assays, the researchers said. Utilizing the MSdPCR methylation data and the SNP data, a balanced bagging classifier model was trained on the Framingham training set, and the final CHD status prediction model was tested on the Framingham test set and externally validated in the two other cohorts. 

In the Framingham test cohort, the area under the curve was .82, sensitivity was .78, and specificity was .74, while in the Intermountain cohort the AUC was .75, sensitivity was .76, and specificity was .71. In the Iowa cohort, the AUC was .88 and sensitivity and specificity were both .82.

The six methylation markers map to at least six distinct, potentially modifiable pathways known to be involved in the pathogenesis of ischemic heart disease, the researchers said, and "most of the power and the sensitivity of the CHD status prediction model is driven by the DNA methylation markers." However, the SNPs "make noticeable contributions to the specificity" through the gene-methylation interaction effects, they added.

The researchers touted the consistent accuracy, sensitivity, and specificity across a broad range of cohorts, as each cohort had diversity in severity of CHD and presentation. The model does not include age or sex as predictive markers, and excluding age and sex didn't affect the performance of the model, the authors noted.

They also said that the model can be further optimized in the future by using a larger cohort with better characterized subjects. The selection of a wide range of CHD patients and the corresponding sensitivity of the model "will not only facilitate that additional optimization but may also facilitate the development of molecular endophenotypes predictive of key CHD features," they said.

Certain markers may help distinguish between obstructive and nonobstructive CHD, allowing patients who may benefit from percutaneous coronary intervention to be prioritized for angiography, they cited as an example. "To further achieve these goals, it will be necessary to recruit and to characterize more subjects whose CHD is robustly characterized using clinical, functional, and diagnostic modalities," they added.

In a statement, Cardio Diagnostics CEO and Cofounder Meesha Dogan said the firm's "commitment is to democratize access to essential cardiac care, particularly in rural areas where advanced diagnostic tools are scarce and cardiovascular specialists are even more rare." 

"PrecisionCHD's remote accessibility and deployment capability in non-specialized settings are pivotal in achieving this goal," she said.