Table 3. Relationship of methylation of leukocytes with other tissues.

	Methylation of leukocytes compared to other tissues (Pearsons r1)
	Aorta	Dura	Trigeminal Caudal Nucleus	Trigeminal Ganglion
Calca	−0.23	−0.33	−0.38	−0.01
Ramp1	0.26	0.64	0.40	0.40
Crcp	0.04	0.33	0.09	0.17
Calcrl	0.08	0.29	0.52	−0.37
Usf2	0.07	0.50	0.07	−0.06
Esr1	0.43	0.53	0.18	0.58
Gper	0.01	0.02	0.17	0.07
Nos3	0.02	0.18	−0.36	−0.38
Mthfr	0.33	−0.09	−0.02	0.13

¹) The relevance of the correlations can best be appreciated from a prediction perspective. Assume that we are interested in estimating the mean amount of methylation of some gene in some tissue. If we have no specific information, the best we can do is take the observed mean in our sample. The uncertainty is quantified by the observed standard deviation (SD). If a linear relationship with the mean amount of methylation in blood exists, a prediction model could be derived. Using basic statistical theory, one can show that the uncertainty is reduced to c * SD, where c follows form the formula c² = 1−r², if r is the correlation coefficient. To get c = 0.5, thus halving the uncertainty, r has to be as high as 0.87. Conversely, when r = 0.5, c = 0.87 and thus only a 13% reduction is obtained. For reliable prediction really high correlations (e.g. r = 0.97 for c = 0.25) are needed.