A LASSO penalized regression approach for genome-wide association analyses using related individuals: application to the Genetic Analysis Workshop 19 simulated data

Table 1 Results of the analyses of the 200 replicates of the simulated data for SBP

Method^a	IC^b	Significant SNPs^c		QTLs discovered^d			SNPs within 2 kbp of QTL^e
Method^a	IC^b	Mean^f	SD^f	TDR^g	Mean^f	SD^f	TDR^g	Mean^f	SD^f
PLINK	–	137.1	98.9	0.020	0.02	0.14	0.240	0.23	0.60
Add	AIC	14.4	5.3	0.000	0.00	0.00	0.010	0.01	0.10
Add	BIC	7.0	3.7	0.000	0.00	0.00	0.010	0.01	0.10
Add-Dom	AIC	14.4	5.3	0.000	0.00	0.00	0.010	0.01	0.10
Add-Dom	BIC	7.1	3.7	0.000	0.00	0.00	0.010	0.01	0.10
Ind	AIC	22.1	9.1	0.000	0.00	0.00	0.020	0.02	0.14
Ind	BIC	15.0	6.2	0.000	0.00	0.00	0.010	0.01	0.10

^aMethod used/covariance structure of random polygenic effect (for LASSO): PLINK PLINK analysis results, Add assuming only additive genetic variance, Add-Dom assuming both additive and dominance genetic effect, Ind assuming members independent
^bInformation criterion used to select best model: AIC Akaike, BIC Bayesian
^cNumber of SNPs with a p value <1E-05 (PLINK) or nonzero coefficient on the optimal LASSO model
^dNumber of actual simulated QTLs with a p value <1E-05 (PLINK) or nonzero coefficient on the optimal LASSO model
^eNumber of SNPs with a p value <1E-05 (PLINK) or nonzero coefficient on the optimal LASSO model located within 2 kbps from a QTL
^fMean and standard deviation of the number of SNPs over the 200 replicates
^gProportion of replicates with at least 1 SNP/QTL with a p value <1E-05 (PLINK)/nonzero coefficient on the optimal LASSO model

ISSN: 1753-6561