Feature (gene, protein, or probe) intensity estimates for each of the assays were pre-processed and normalised as described in the individual assay sections and subsequently transformed into a standard normal distribution across lines. For each feature in each assay, variance was partitioned using a linear mixed model (implemented in the lme4 R package) fitted with all metadata variables as random effects. Only lines with complete metadata information were included in each one of the analyses, these numbers are shown in parenthesis in Fig. 3a-c. The variance components were normalized to sum to one and subsequently averaged across the different sets of features considered (Fig. 3). The fraction of non-technical variance explained by each biological or experimental factor refers to the variance explained by the factor, divided by the total variance minus the variance explained by assay batches (see below for a definition of experimental and assay batch factors). Confidence intervals for the Cellomics variance components were obtained with the ‘profile’ method implemented in the ‘confint’ function of the lme4 package. The following random effects were included for each assay: Methylation
