Fig 1.
Workflow of the regression methods.
Fig 2.
PCA from the four different pre-processing approaches.
Each color corresponds to a different batch. A = per batch normalization with no QCs, B = per batch normalization with QC correction, C = merged normalization with no QCs, D = merged normalization with QC correction.
Table 1.
Association between the first five Principal Components (PC) and the batch effect using R2 before and after correcting for batch effect.
We underlined the highest correlation value per row.
Table 2.
Number of true positives (TP) and false positives (FP) found in the different simulations: with and without QCs for the different effect sizes.
Table 3.
Number of TP and FP found in the different simulations after adding random error to the original simulation.
Table 4.
Number of TP and FP found in the different simulations for the reduced dataset derived from the original population.
Table 5.
Number of TP and FP found in the different simulations for the unbalanced study design dataset.