Skip to main content
Fig. 4 | BMC Genomics

Fig. 4

From: Protein complex-based analysis is resistant to the obfuscating consequences of batch effects --- a case study in clinical proteomics

Fig. 4

a 3D-Principal Component Analysis: Class effects dominate in the simulated data, given the removal of PC1 (Two examples are shown: D2.2.301 and D2.2.302). b Heatmaps and hierarchical clustering (HCL): The remaining PCs may also be used as individual variables for clustering, and provide strong discrimination between classes D and D*. c Combining two datasets with different batch effects: Datasets A and B have the same differential feature set but different batch effects. Combining these followed by analysis of all principal components (PC) shows batch effects dominate. However, removal of PC1 perfectly recovers class-effect discrimination without having to perform any feature selection (Notation: A/B_D/D*_1/2 refers to the dataset, class and batches respectively)

Back to article page

BMC Genomics

ISSN: 1471-2164

Contact us