Batch effect correction: Input data format for removeBatchEffect from limma package

Question

[mohammedtoufiq91]

Hello,

I am working with the Biomark HD Fluidigm high throughput qPCR transcriptomics datasets (consists of 240 target genes and 8 housekeeping genes). The dataset consists of 250 samples processed in 5 different batches (50 samples/batch). Dataset of each batch was normalized to calculate Delta Ct (Difference between the Target gene and Geometric mean of 8 housekeeping gene (i.e GM_HKgs)). Then, transformed to - Delta Ct (negative delta Ct) because we want to apply downstream analyses like ANOVA to log-scale normalized relative expression values, as just using -1 * Delta Ct as readout is quite valid too. I understand from various posts removebatcheffects (limma package) would be best to use. My question is what data type should be used as input, Delta Ct or Negative Delta Ct (-1*Delta Ct).

Answer 1

[Gordon Smyth]

You could use negative delta ct, which would be fine, or see https://support.bioconductor.org/p/67032/ for a more sophisticated normalization method that we use for qPCR.

You don't necessarily need removeBatchEffect however. The limma DE/anova analysis of qPCR ct values does not use removeBatchEffects.

[mohammedtoufiq91]

Hi Gordon Smyth Thank you for the inputs and suggestions. I will also try to run the limma DE for the qPCR Ct values and benchmark the data.

Meanwhile, if I would like to use removeBatchEffect for continuity of the qPCR data analysis. Can I still use the same? Additionally, the resultant matrix (y2) generated from this function could be used for further downstream analysis like t-test/ANOVA/plots or just utilized only for the visualization purposes like PCA and Heatmap?

y2 <- removeBatchEffect(y, batch)

[Gordon Smyth]

I'm not sure what you mean by "continuity". I recommend that removeBatchEffect only be used for visualization, as is said on the help page for that function.

[mohammedtoufiq91]

Hi Gordon Smyth Noted. I will use removeBatchEffect on my Negative Delta Ct values for the visualization only. Thank you very much.