Fig. 1.

a Overview of DNA methylation analysis steps and commonly used options. We identified four steps in the procedure which often vary in literature: (A) DNAm value preprocessing, (B) statistical test, (C) cell count correction, (D) hidden confounder correction. We selected one combination of options and then varied these a single step at the time. These models were applied to age, BMI, and smoking. Each model was meta-analyzed in each combination of three discovery and one replication cohorts. The average replication rate and number of replicated genes of these four analyses were used to evaluate strategies. The base model is connected by the black line and includes Beta-3IQR dataset, an LM model, measured cell count correction, known technical confounder correction (TCs) (plate and row) and applying Bonferroni correction. HCs, hidden confounders, calculated after regressing out technical covariates (2), cell counts (3) or both (4). b Overview of gene expression analysis steps and commonly used options. We identified four steps in the procedure which often vary in literature: (A) normalization, (B) expression, (C) tests, and (D) technical covariates. We selected one combination of options and then varied these a single step at the time. These models were applied to age, BMI, and smoking. Each model was meta-analyzed in each combination of three discovery and one replication cohorts. The average replication rate and number of replicated genes of these four analyses were used to evaluate strategies. The base model is connected by the black line; Voom normalization, including all genes, a LM for statistical analysis, including technical covariates and cell counts and applying Bonferroni correction