[Preprint]. 2023 Feb 14:2023.02.10.23285764. [Version 1] doi: 10.1101/2023.02.10.23285764

Table 2.

Estimated contributions of noteworthy CpG sites on the mediation pathway between low education and HbA1c

CpG Name	Chromosome	Nearby Gene(s)	USCS RefGene Group	Univariate (0 sites identified)	HIMA (3 sites identified)	HDMA (11 sites identified)	MedFix (3 sites identified)	Pathway LASSO (141 sites identified)	BSLMM (3 sites identified)
cg10508317	17	SOCS3	Body	3.48 x10⁻²	1.59 x10⁻²^*	3.56 x10⁻²^*	2.90 x10⁻²^*	2.35 x10⁻²^*	0.25 x10⁻²
cg01288337	14	RIN3	Body	3.35 x10⁻²	1.47 x10⁻²^*	2.82 x10⁻²^*	2.70 x10⁻²^*	4.43 x10⁻²^*	0.21 x10⁻²
cg10244976	16	LMF1	Body	3.00 x10⁻²	0	2.78 x10⁻²^*	0	2.23 x10⁻²^*	0.19 x10⁻²
cg07516252	14	REC8	TSS200	2.72 x10⁻²	0	2.24 x10⁻²^*	0	2.26 x10⁻²^*	0.26 x10⁻²
cg07571519	10	C10orf105; CDH23	3'UTR; Body	2.53 x10⁻²	0.33 x10⁻²^*	3.67 x10⁻²^*	1.47 x10⁻²^*	2.81 x10⁻²^*	0.21 x10⁻²
cg23079012	2	LINC00299	Body	2.27 x10⁻²	0	1.99 x10⁻²^*	0	1.98 x10⁻²^*	0.29 x10⁻²
cg01587454	8	DCAF4L2	1stExon	1.77 x10⁻²	0	2.10 x10⁻²^*	0	1.99 x10⁻²^*	0.38 x10⁻²
cg27527503	4	HADH	TSS1500	1.75 x10⁻²	0	1.86 x10⁻²^*	0	1.27 x10⁻²^*	0.23 x10⁻²
cg25891647	11	GRAMD1B	Body	−1.27 x10⁻²	0	−3.42 x10⁻²^*	0	−3.02 x10⁻²^*	−0.33 x10⁻²
cg08473752	17	NLK	Body	−0.70 x10⁻²	0	−2.34 x10⁻²^*	0	−2.32 x10⁻²^*	−0.22 x10⁻²
cg12644059	15	BLM	N/A¹	−0.03 x10⁻²	0	−2.31 x10⁻²^*	0	−1.84 x10⁻²^*	−0.22 x10⁻²

Selected as noteworthy by given method

CpG site cg12644059 is 3.240kb from the final base pair of the BLM gene

Table includes all CpG sites that were selected as having a noteworthy mediation contribution by at least two of the implemented methods out of 2,000 CpG sites in total. Criteria for CpG identification varied by method. All estimates are adjusted for age, sex, race, and the estimated proportions of residual non-monocytes as fixed effects, along with methylation chip and position as random effects to address potential batch effects. Note that for HIMA, HDMA, MedFix, and pathway LASSO, which fit high-dimensional regression models, we used additional pre-screening to reduce the number of mediators in advance to only n/log(n) ≈ 141 CpG sites, which is the approach recommended by the HIMA and HDMA authors and helps with statistical and computational efficiency (see Methods). Pathway LASSO selected all of these 141.