Batch effects are systematic technical variation between experimental batches. It can dominate biological signal. To address them:
- (Design) This is a prerequisite - ensure balanced experimental design (don’t confound treatment with batch);
- (Detect) Check PCA plots to see if the sample clustering are associated with batch labels;
- ProteoAnalyst offers two types of approaches
- (Remove) Use ComBat for batch correction before performing statistical analysis;
- (Model) Some methods such as linear modeling (limma) can consider batch effect as covariates during data analysis