Prediction of metabolic syndrome using machine learning approaches based on genetic and nutritional factors: a 14-year prospective-based cohort study

Shin, Dayeon

doi:10.1186/s12920-024-01998-1

Table 5 Results on predictive performance metrics for metabolic syndrome prediction analysis using machine learning algorithms based on epidemiological, nutritional and genome-wide polygenic risk score (gPRS)

From: Prediction of metabolic syndrome using machine learning approaches based on genetic and nutritional factors: a 14-year prospective-based cohort study

Model	Algorithm	N	Event N	Pred N	AUC (95% CIs)	p-value (H₀: AUC = 0.5)	Sensitivity	Specificity	PPV	NPV	Cutoff	DeLong p-value	F1-score
Model 1 = Epidemiological factors + dried laver	Cox	1690	1430	1419	0.808 (0.779–0.837)	< 0.001	0.760	0.708	0.839	0.114	-0.285	Ref.	0.798
	DNN	1690	1430	1017	0.864 (0.843–0.885)	< 0.001	0.693	0.896	0.973	0.346	0.434	< 0.001	0.809
	SVM	1690	1430	1231	0.879 (0.858–0.900)	< 0.001	0.827	0.792	0.82	0.083	-0.742	< 0.001	0.823
	SGD	1690	1430	889	0.653 (0.619–0.688)	< 0.001	0.569	0.712	0.916	0.231	0.443	< 0.001	0.702
	RAF	1690	1430	1437	0.989 (0.980–0.998)	< 0.001	1.000	0.973	0.995	1	0.444	< 0.001	0.997
	NBA	1690	1430	1156	0.784 (0.752–0.816)	< 0.001	0.756	0.715	0.935	0.346	0.077	0.021	0.836
	ADB	1690	1430	1136	0.847 (0.823–0.872)	< 0.001	0.754	0.777	0.949	0.365	0.336	< 0.001	0.840
Model 2 = Epidemiological factors + dried laver + genome-wide polygenic risk score (gPRS) Z-score 1	Cox	1690	1430	1369	0.826 (0.799–0.853)	< 0.001	0.838	0.669	0.84	0.128	-0.458	Ref.	0.839
	DNN	1690	1430	1144	0.881 (0.861–0.902)	< 0.001	0.777	0.873	0.971	0.416	0.362	< 0.001	0.863
	SVM	1690	1430	1310	0.885 (0.864–0.905)	< 0.001	0.800	0.835	0.824	0.079	-0.608	< 0.001	0.812
	SGD	1690	1430	1108	0.701 (0.667–0.734)	< 0.001	0.706	0.612	0.909	0.273	0.374	< 0.001	0.795
	RAF	1690	1430	1437	0.990 (0.981–0.999)	< 0.001	1.000	0.973	0.995	1	0.448	< 0.001	0.997
	NBA	1690	1430	1126	0.797 (0.766–0.828)	< 0.001	0.740	0.742	0.94	0.34	0.084	0.005	0.828
	ADB	1690	1430	1124	0.869 (0.846–0.891)	< 0.001	0.755	0.823	0.959	0.378	0.358	< 0.001	0.845
Model 3 = Epidemiological factors + dried laver + gPRS Z-score 2	Cox	1690	1430	1585	0.867 (0.843–0.891)	< 0.001	0.816	0.773	0.842	0.095	-0.191	Ref.	0.829
	DNN	1690	1430	1258	0.910 (0.893–0.927)	< 0.001	0.857	0.877	0.974	0.525	0.350	< 0.001	0.912
	SVM	1690	1430	1376	0.907 (0.888–0.926)	< 0.001	0.864	0.827	0.831	0.089	-0.699	< 0.001	0.847
	SGD	1690	1430	1152	0.802 (0.772–0.832)	< 0.001	0.753	0.708	0.934	0.342	0.452	< 0.001	0.834
	RAF	1690	1430	1437	0.987 (0.977–0.998)	< 0.001	1.000	0.973	0.995	1	0.470	< 0.001	0.997
	NBA	1690	1430	1228	0.839 (0.811–0.867)	< 0.001	0.811	0.750	0.946	0.42	0.071	0.005	0.873
	ADB	1690	1430	1204	0.881 (0.857–0.904)	< 0.001	0.804	0.788	0.954	0.422	0.323	0.017	0.873
Model 4 = Epidemiological factors + dried laver + gPRS Z-score 3	Cox	1690	1430	1320	0.966 (0.954–0.979)	< 0.001	0.909	0.923	0.985	0.649	0.225	Ref.	0.945
	DNN	1690	1430	1309	0.960 (0.946–0.974)	< 0.001	0.909	0.965	0.993	0.659	0.390	0.094	0.949
	SVM	1690	1430	1556	0.975 (0.965–0.986)	< 0.001	0.923	0.950	0.836	0.037	-0.538	0.002	0.877
	SGD	1690	1430	1339	0.958 (0.944–0.972)	< 0.001	0.917	0.896	0.98	0.664	0.470	< 0.001	0.947
	RAF	1690	1430	1437	0.990 (0.981–0.999)	< 0.001	1.000	0.973	0.995	1	0.454	< 0.001	0.997
	NBA	1690	1430	1339	0.944 (0.927–0.962)	< 0.001	0.914	0.873	0.975	0.647	0.045	0.002	0.944
	ADB	1690	1430	1387	0.977 (0.966–0.987)	< 0.001	0.954	0.912	0.983	0.782	0.171	< 0.001	0.968
Model 5 = Epidemiological factors + dried laver + gPRS Z-score 4	Cox	1690	1430	1397	0.987 (0.978–0.996)	< 0.001	0.967	0.950	0.99	0.84	0.087	Ref.	0.978
	DNN	1690	1430	1386	0.985 (0.975–0.995)	< 0.001	0.965	0.977	0.996	0.836	0.419	0.237	0.980
	SVM	1690	1430	1591	0.989 (0.980–0.998)	< 0.001	0.986	0.958	0.84	0.061	-0.998	0.428	0.907
	SGD	1690	1430	1384	0.980 (0.970–0.991)	< 0.001	0.957	0.946	0.989	0.801	0.423	0.001	0.973
	RAF	1690	1430	1437	0.993 (0.985–1.000)	< 0.001	1.000	0.973	0.995	1	0.466	0.082	0.997
	NBA	1690	1430	1391	0.973 (0.960–0.986)	< 0.001	0.962	0.935	0.988	0.813	0.035	0.008	0.975
	ADB	1690	1430	1410	0.994 (0.987–1.000)	< 0.001	0.982	0.977	0.996	0.907	0.318	0.016	0.989
Model 6 = Epidemiological factors + dried laver + gPRS Z-score 5	Cox	1690	1430	1405	0.991 (0.983–0.999)	< 0.001	0.976	0.965	0.994	0.881	0.132	Ref.	0.985
	DNN	1690	1430	1396	0.989 (0.980–0.998)	< 0.001	0.971	0.973	0.995	0.861	0.425	0.105	0.983
	SVM	1690	1430	1610	0.992 (0.983–1.000)	< 0.001	0.987	0.969	0.841	0.05	-0.826	0.713	0.908
	SGD	1690	1430	1360	0.984 (0.975–0.992)	< 0.001	0.946	0.969	0.994	0.764	0.455	< 0.001	0.969
	RAF	1690	1430	1437	0.994 (0.986–1.000)	< 0.001	1.000	0.973	0.995	1	0.445	0.150	0.997
	NBA	1690	1430	1415	0.980 (0.970–0.991)	< 0.001	0.978	0.938	0.989	0.887	0.019	0.013	0.983
	ADB	1690	1430	1440	0.993 (0.986–1.000)	< 0.001	1.000	0.962	0.993	1	0.075	0.393	0.996
Model 7 = Epidemiological factors + dried laver + gPRS Z-score 6	Cox	1690	1430	1415	0.991 (0.984–0.999)	< 0.001	0.984	0.969	0.994	0.916	0.030	Ref.	0.989
	DNN	1690	1430	1690	0.991 (0.984–0.999)	< 0.001	0.969	0.985	0.846	N/A	0.000	0.929	0.903
	SVM	1690	1430	1633	0.991 (0.982–1.000)	< 0.001	0.985	0.977	0.841	0.018	-0.738	0.998	0.907
	SGD	1690	1430	1408	0.987 (0.978–0.996)	< 0.001	0.978	0.965	0.994	0.89	0.406	0.005	0.986
	RAF	1690	1430	1433	0.992 (0.982–1.000)	< 0.001	0.998	0.977	0.996	0.988	0.615	0.986	0.997
	NBA	1690	1430	1422	0.984 (0.974–0.993)	< 0.001	0.986	0.946	0.99	0.918	0.008	0.024	0.988
	ADB	1690	1430	1434	0.992 (0.985–1.000)	< 0.001	0.999	0.977	0.996	0.992	0.353	0.683	0.997
Model 8 = Epidemiological factors + dried laver + gPRS Z-score 7	Cox	1690	1430	1396	0.992 (0.985–0.999)	< 0.001	0.973	0.981	0.996	0.867	0.208	Ref.	0.984
	DNN	1690	1430	1690	0.991 (0.983–0.999)	< 0.001	0.978	0.977	0.846	N/A	0.000	0.342	0.907
	SVM	1690	1430	1633	0.991 (0.982–1.000)	< 0.001	0.979	0.981	0.842	0.035	-0.789	0.411	0.905
	SGD	1690	1430	1408	0.988 (0.981–0.996)	< 0.001	0.978	0.965	0.994	0.89	0.367	0.005	0.986
	RAF	1690	1430	1434	0.994 (0.985–1.000)	< 0.001	0.999	0.977	0.996	0.992	0.615	0.306	0.997
	NBA	1690	1430	1412	0.984 (0.975–0.993)	< 0.001	0.978	0.950	0.991	0.888	0.009	0.023	0.984
	ADB	1690	1430	1431	0.994 (0.986–1.000)	< 0.001	0.997	0.977	0.996	0.981	0.284	0.103	0.996

Epidemiological factors include sex, age, alcohol intake, energy intake, marital status, education status, income status, and smoking status.
Cox, Cox multivariable regression; DNN, deep neural network; SVM, support vector machine; SGD, stochastic gradient descent; RAF, random forest; NBA: Naïve Bayes classifier; ADB, AdaBoost; AUC, area under the ROC curve; PPV, positive predictive value; NPV, negative predictive value; gPRS, genome-wide polygenic risk score.
gPRS Z-score 1 based on p-value < 0.0001, gPRS Z-score 2 based on p-value < 0.001, gPRS Z-score 3 based on p-value < 0.01, gPRS Z-score 4 based on p-value < 0.05, gPRS Z-score 5 based on p-value < 0.1, gPRS Z-score 6 based on p-value < 0.2, and gPRS Z-score 7 based on p-value < 1.0.

Back to article page

ISSN: 1755-8794

Contact us

General enquiries: journalsubmissions@springernature.com

BMC Medical Genomics

Contact us