Resampling-based tests for Lasso in genome-wide association studies

BMC Genomic Data

Table 5 Alcohol Consumption GWAS

SNP	Chr.	Gene	Distance from Gene (bp)	SMA	Lasso-PL	Lasso-AL	Lasso-Ayers
rs7574612	2	LHCGR	-24421	9.4∗10⁻⁶ (S)	S	S	S
rs4836266	5	GRAMD3	-235	1.3∗10⁻⁵ (N)	S	N	N
rs211598	6	EYA4	-29329	9.3∗10⁻⁶ (S)	S	S	S
rs4385434	8	hCG_1814486	-129857	5.5∗10⁻⁶ (S)	S	S	S
rs7136989	12	GOLGA3	-244	1.7∗10⁻⁵ (N)	S	N	N
rs6072694	20	PTPRT	-1180	2.1∗10⁻⁶ (S)	S	S	S
rs233278	21	KRTAP10-4	-1677	1.3∗10⁻⁶ (N)	S	N	N

For Tables 5 and 6: “S” means ”significant” and “N” means “not significant” using significance level α=1.18∗10⁻⁵. Note that Lasso-PL, Lasso-AL, and Lasso-Ayers cannot provide exact p-values, but selects significant SNPs while attempting to control the type-1-error rate at level α. One could fit multiple penalized regression models and estimate λ that controls the type-1-error rate at various orders of magnitude (e.g. 10⁻⁵,10⁻⁶, etc) to get a better idea of the significance of each selected SNP (not done here)

ISSN: 2730-6844