line. Among the DNA samples in the ‘other’ category, 77% are from cell lines, whereas the values for alcohol cases and controls are 34% and 25%, respectively. The missing call rate for cell lines is very significantly higher than for blood (p<9 × 10−14), as shown in Figure S9 (0.12% versus 0.10%). Therefore, allelic frequency differences between ‘other’ versus ‘case’ and ‘control’ categories are potentially biased by nonrandom missingness. In a situation like this, it may be useful to adjust for tissue type in the association analysis, by including it as a regression covariate.
