Fig. 2

An overview of the voom-based mean-variance modeling methods applied. On the PBMC1 pseudo-bulk data, group-specific weights estimated using voomQWB by defining each block (group) as different levels of the symptom variable are plotted in (a). The equal weight (\(=1\)) level is plotted as a dashed line. Across all observations, gene-wise square-root residual standard deviations are plotted against average log-counts in gray in (b). voom applies a LOWESS trend (black curve) to capture the relationship between the gene-wise means and variances. Based on the final precision weights used in voomQWB, adjusted curves for each block are plotted in (c), where replicates in the same group share the same curve. Different colors and line types represent different groups (blocks). Dashed lines were used to avoid over-plotting. When voomByGroup is used, LOWESS trends are fitted separately to the data from individual groups to capture any distinct mean-variance trends that may be present (d–f). All group-specific trends from this dataset are plotted together in panel g, with different colors per group