Revisiting genetic artifacts on DNA methylation microarrays exposes novel biological implications

Table 1 Benchmarking of UMtools on sex chromosome- and SNP-targeting probes

UMtools
Tool	Scale	Purpose
U/M plot	Targeted	Cluster visualization
bGMM	Targeted	Cluster assignment for a target number of clusters
BC(CV)	Epigenome-wide	Ambivalence in noise-to-signal ratio detection
cor_MZ(CV)	Epigenome-wide	Genetic control for noise-to-signal ratio
K-caller	Epigenome-wide	Cluster counting
Benchmarking
Markers		ChrY	ChrX_inact	ChrX_{hypermeth + escape}	SNP probes
# probes		266	3,981	3,028	65
Expected K		2	2	1 (large n)	3
Probe failure in females		Yes	No	No	No
bGMM (K = 2 or 3)	Twin cluster assignment agreement	0.994	0.991	0.479^a	0.997
BC(CV_logT) and cor_MZ(CV)	Genetics-related probe failure	0.951	0.001	0.001	0.000
K-calling	Correct # clusters predicted	0.977	0.902	0.999	1.000

^a Full separation between males and females is not observed in a large cohort as E-risk (Fig. 1D); it can be seen though in smaller datasets (Additional file 1: Fig S2A) that are not so strongly affected by batch effects

ISSN: 1474-760X