Key result
Combining clinical and genetic predictors in machine learning improves obstructive CAD prediction over clinical-only models.
Why the study?
Developing and clinically validating machine learning algorithms integrating genetic biomarkers and clinical features may improve diagnostic accuracy for CAD and avoid unnecessary invasive procedures.
Does a machine learning model integrating genetic and clinical predictors improve the prediction of coronary artery disease existence and severity compared to clinical predictors alone?
Observational (n=953)
Single-blind
Yes
Does a machine learning model integrating genetic and clinical predictors improve the prediction of coronary artery disease existence and severity compared to clinical predictors alone?
Absolute Event Rate: 0.798% vs 0.757%
p-value: p=0.014
Integrating genomic profiling with clinical data using machine learning improves the diagnostic prediction of coronary artery disease existence and severity, potentially reducing the need for invasive procedures.
Cardiovascular diseases (CVDs) present multifactorial pathophysiology and produce immense health and economic burdens globally. The most common type, coronary artery disease (CAD), shows a complex etiology with multiple genetic variants to interplay with various clinical features and demographic traits affecting CAD risk and severity. The development and clinical validation of machine learning (ML) algorithms that integrate genetic biomarkers and clinical features can improve diagnostic accuracy for CAD avoiding, thereby, unnecessary invasive procedures. To this end, we present, here, the development of a data-driven ML approach able to predict the existence and severity of CAD based on the analysis of 228 single nucleotide polymorphisms (SNPs) and clinical and demographic data of 953 patients enrolled in the Genetic Syntax Score (GESS) trial ( NCT03150680 ). Two competing ensemble models (one with clinical predictors and another with clinical plus genetic predictors) were built and evaluated to infer their prediction capabilities. The ensemble model with both clinical and genetic predictors exhibited superior diagnostic performance compared to the competing model with only clinical predictors. The proposed ML framework identified a total of eight contributing SNPs as predictors for the existence of obstructive CAD and seven significant SNPs for the severity of CAD. Such algorithms positively contributes to global efforts aiming to predict the risk and severity of CAD in early stages, thus lowering the cost as well as achieving prognostic, diagnostic, and therapeutic benefits in healthcare and improving patient outcomes in a non-invasive way. Overall, the design and execution of this trial reinforces clinical decision-making and facilitate the harmonization in digitized healthcare within the concept of precision medicine. Clinical Trial Registration NCT03150680; https://clinicaltrials.gov/study/NCT03150680?cond=NCT03150680&rank=1
No takes yet. Share an insight, caveat, or question.
Chatzopoulou et al. (2024) conducted an observational in Coronary artery disease (n=953). Machine learning model with clinical and genetic predictors (Model B) vs. Machine learning model with clinical predictors only (Model A) was evaluated on Prediction of obstructive CAD (AUC) (p=0.014). A machine learning model combining clinical and genetic predictors predicted obstructive coronary artery disease better than a clinical-only model (AUC 0.798 vs 0.757, p=0.014).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: