1. Knowledge Hub
  2. Methods
  3. NanoString nCounter Gene Expression

Covariate correction for NanoString nCounter Gene Expression data

Similar to nSolver Advanced Analysis, you can eliminate confounding and undesirable "extra" experimental variables in Gene Expression experiments with covariate correction. ROSALIND intelligently determines which experiment attributes may be valid covariates, reducing complexity for this advanced capability. Every NanoString nCounter Gene Expression experiment has free access to covariate correction. Add covariate correction when setting up a comparison within an experiment to see Venn Diagrams and covariate Meta-Analyses for rapid interpretation of corrected and uncorrected values.


For example, an experiment evaluating two treatments across three cell lines may benefit from covariate correction to eliminate confounding differences between cell lines. Without covariate correction, six comparisons for each treatment and cell line would be required to discover meaningful insights in differential expression. Using covariate correction, a single comparison enables the exploration of the treatment and control across all three cell lines, simultaneously.

Methods used for raw data (RCC files):

Effects of confounding variables or covariates can be optionally assessed with covariate correction, which is performed using the generalized linear model developed by the NanoString Biostatistics team. Detailed information can be found in the nSolver Advanced Analysis 2.0 User Manual pages 50-52. Covariate (also called confounder) correction is described specifically on page 52 under the section titled "Variables". 


Methods used for normalized counts (CSV file):

The limma R library (1) is used to calculate covariate correction for NanoString Gene Expression experiments created from normalized counts.


  1. Ritchie, M. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res, 43(7) (2015).