Polygenic risk scores outperform machine learning methods in predicting coronary artery disease status

Abstract

Coronary artery disease (CAD) is the leading global cause of mortality and has substantial heritability with a polygenic architecture. Recent approaches of risk prediction were based on polygenic risk scores (PRS) not taking possible nonlinear effects into account and restricted in that they focused on genetic loci associated with CAD, only. We benchmarked PRS, (penalized) logistic regression, naïve Bayes (NB), random forests (RF), support vector machines (SVM), and gradient boosting (GB) on a data set of 7,736 CAD cases and 6,774 controls from Germany to identify the algorithms for most accurate classification of CAD status. The final models were tested on an independent data set from Germany (527 CAD cases and 473 controls). We found PRS to be the best algorithm, yielding an area under the receiver operating curve (AUC) of 0.92 (95% CI [0.90, 0.95], 50,633 loci) in the German test data. NB and SVM (AUC ~ 0.81) performed better than RF and GB (AUC ~ 0.75). We conclude that using PRS to predict CAD is superior to machine learning methods.

Current Developments of Clinical Sequencing and the Clinical Utility of Polygenic Risk Scores in Inflammatory Diseases

“…To assess this question for CAD, Gola et al. ( 63 ) compared various methods from the field of machine learning (ML), which offer attractive algorithms to model non-linear effects, with a GPRS in a case-control data set of samples of European descent from the German population. It turned out that a simple GPRS outperformed all other algorithms under consideration by means of a nested cross-validation.…”

A Rule-Based Expert System to Assess Coronary Artery Disease Under Uncertainty

“…In [ 28 ], Multi-Criteria Decision Making Methods were presented for accessing CAD under uncertainty where presence and absence of CAD is predicted through using symptom and signs of CAD. But these approaches report neither the number of blocked arteries nor the significance of severity of the disease [ 8 , 16 , 26 , 28 , 29 ]. Weak parameters, like signs and symptoms, are used for predicting CAD as well as for predicting the similar types of diseases like mitral regurgitation, dilated cardiomyopathy, congenital heart disease, hyper-tropic cardiomyopathy, myocardial infarction etc.…”

“…Some researchers developed the Medical Decision Support System (MDSS) to predict CAD. Other proposer polygenic risk scores (PRS), a nonlinear, for CAD prediction with accuracy an 0.92 under the receiver operating curve (AUC) [ 8 ].…”

“…Moreover, walking on a trade mill in a stress test makes the patient discomfort heart function than normal condition. Nowadays, support vector machine (SVM) [ 7 , 8 ] and artificial neural network (ANN) [ 5 , 8 – 17 ] based Clinical Decision Support Systems (CDSS) [ 18 – 20 ] are developed for CAD prediction. Unfortunately, SVM and ANN have no direct impact on the reasoning process due to their black-box- type modeling approaches.…”

Interaction-based feature selection algorithm outperforms polygenic risk score in predicting Parkinson’s Disease status

“…Nevertheless, in a recent study a PRS model outperformed the five machine learning algorithms Naïve Bayes classifier, regularized regression, random forest, gradient boost, and support vector machine used to build prediction models for coronary artery diseases status. (Gola et al, 2020) Here, we explore the potential of using contextual information obtained via data mining to strongly reduce the hypothesis space, which, in turn, allows for testing a small set of complex hypotheses, containing interaction of multiple variants. This approach organizes data mined from journal articles, pathway libraries, protein co-expression libraries and drug candidate libraries into a hierarchical graph, which generates disease-specific hypotheses based on interactions of genetic variants.…”