Table 1 Statistical association of technical factors or biological signals before and after data adjustment.

From: Controlling technical variation amongst 6693 patient microarrays of the randomized MINDACT trial

Association

Unadjusted dataset

Adjusted dataset

(A)

Canonical correlation of technical factors with PC1–4

Automated labeling/PC1–4

0.66

0

Labeling yield/PC1–4

0.58

0.27

Reagent × reference RNA/PC1–4

0.53

0

(B)

Pearson correlation of biological signals–associations with TargetPrint (TP) readout

ER/TP-ER

0.98

0.95

HER2/TP-HER2

0.79

0.80

PC1/TP-ER

0.31

0.80

PC1/TP-HER2

0.09

0.16

  1. Part A lists the canonical correlation of technical factors with the top four principal components (PC1–4) before and after data adjustment. Canonical correlations are a multivariate generalization of Pearson correlation that finds the strongest possible association between two sets of variables (between linear combinations of the variables in each set). The change in statistical association between the unadjusted and adjusted data shows the disappearance or decrease of the effect of technical factors on gene expression as a result of the data adjustment.
  2. Part B gives the correlation of gene expression with ER and HER2 status that was used to assess whether biological signals were preserved after adjustment. TargetPrint (TP)-scores of the unadjusted dataset was used as a suitable substitute for the immunohistochemistry status that was not available to us at the time of analyses. The absolute Pearson correlations between ER and HER2 gene expression signals and their respective TargetPrint scores (ER/TP-ER and HER2/TP-HER2, respectively) are shown for the unadjusted and adjusted datasets. Additionally, the absolute Pearson correlations between the first principal component (PC1) and TargetPrint scores for ER and HER2 (PC1/TP-ER and PC1/TP-HER2, respectively) are shown for the unadjusted and adjusted datasets.
  3. p-values are non-informative (see “Methods” for more information) and therefore not given.
  4. ER estrogen receptor, HER2 erb-b2 receptor tyrosine kinase 2, PC principal component, TP-ER TargetPrint for ER, TP-HER2 TargetPrint for HER2.
Back to article page