Fig. 1

It is critical to control for smoking in an epigenome-wide association study (EWAS) of schizophrenia. Manhattan plots comparing association P-values (y-axis, −log10 scale) against genomic location (x-axis) for (a) an EWAS of schizophrenia case–control status without inclusion of a smoking covariate, (b) a published EWAS of tobacco smoking (never versus current) [22], and (c) an EWAS of schizophrenia case–control status including a DNA methylation-derived smoking score as a covariate. There is considerable overlap between the associated loci in panels a and b (e.g., on chromosomes 2, 5, 6, and 19), implying that the elevated rate and intensity of smoking in patients with schizophrenia [25–27] is a severe confounder in this analysis. In contrast, there is no overlap in the associated loci shown in panels b and c, suggesting that the derived smoking score effectively removes these false positives