Table 2.

DNA methylation-based biomarkers of exposure to smoking.

Study	Timing of:		Persistence	Tissue	Model	Predictive accuracy	Notes
	Smoking exposure	DNAm measure
Reese et al (2017)[72]	Prenatal	Birth	Days-months	Cord blood	Score: linear combination of 28 loci and LASSO regression coefficients	AUC1 = 0.90
Ladd-Acosta et al (2016)[31]		Early Childhood (Age 5)	5 years	Peripheral blood	SVM classifier: 26 loci	AUC1 = 0.87
Richmond et al (2018)[52]		Adulthood (Age 30)	~30 years	Peripheral blood	Score: weighted† sum of 568 loci*	AUC1 = 0.69
		Adulthood (Age 30)	~30 years	Peripheral blood	Score: weighted† sum of 19** loci	AUC1 = 0.72
		Adulthood (Age 30)		Peripheral blood	Score: weighted† sum of 2623*** loci	AUC1 = 0.57	Shows current personal smoking score is not a good predictor of prenatal smoking exposure.
Shenker etl (2013)[73]	Adult	Later Adulthood		Peripheral blood	GLM using 4 loci	AUC2 = 0.83 AUC3 = 0.97	DNAm biomarker better predictor than cotinine (AUC=0.47) DNAm correlated with time since quitting and duration of smoking

DNAm, DNA methylation

GLM, generalized linear model

SVM, support vector machine

¹Predicting prenatally exposed versus unexposed

²Predicting former vs never smokers

³Predicting current versus never smokers

^*568 DNAm loci originally identified in cord blood

^**DNAm loci originally identified in blood during middle childhood

^***DNAm loci identified in adult blood as predictive of current smoking

^†Weights determined by: per CpG effect size/average effect size for all measured CpG sites.