Methylation data imputation performances under different representations and missingness patterns

Results: We make an extensive analysis of the imputation performances of seven imputation methods on simulatedmissing completely at random(MCAR),missing at random(MAR) andmissing not at random(MNAR) methylation data. We further consider imputation performances on the popular beta- andM-value representations of methylation levels. Overall,beta-values enable better imputation performances thanM-values. Imputation accuracy is lower for mid-range beta-values, while it is generally more accurate for values at the extremes of the beta-value range. The MAR values distribution is on the average more dense in the mid-range in comparison to the expected beta-value distribution. As a consequence, MAR values are on average harder to impute.