Skip to main content
. 2020 Apr 22;12:58. doi: 10.1186/s13148-020-00850-4

Table 5.

Origins of smoking DNAm scores employed in the current analysis

Phenotype Origin publication EWAS model # CpG sites
Smoking ‘Epigenetic Signatures of Cigarette Smoking’ Joehanes et al. [19] Linear mixed models were conducted, then combined in a random-effects model meta-analysis (450 K). After meta-analysis, one set of CpGs was selected based on a Bonferroni P value of P < 1 × 10−7 (485,381 tests) and another was selected based on a genome-wide false discovery rate P value < 0.05. Bonferroni model: 2623, FDR model: 18760
‘Self-reported smoking, serum cotinine, and blood DNA methylation’ Zhang et al. [20] An EWAS (450 K) of cotinine concentration was conducted using median quantile regression, then CpG sites were individually validated against estimated average cigarettes per day using restricted cubic spline regression. Results were filtered by optimising AUCs derived from logistic regression for smoking status (current vs never; former vs never). 4
‘Bayesian reassessment of the epigenetic architecture of complex traits’ Trejo Banos et al. [17] EWAS (MethylationEPIC) were conducted using a Bayesian framework. 59
‘Epigenetic prediction of complex traits and death’ McCartney et al. [4] EWAS (MethylationEPIC) were conducted using a LASSO regression model with k-fold (k = 10) cross-validation. 233