Values EXCEEDS the original data range after batch effect correction.

Question

[Mary]

Hi

I have gene coverage (%) data in a matrix structure. since the coverage estimated is in percentage, it falls in range of 0-100%. now I ran limma to remove batch effect, and see that the range of gene coverage is now 0-6000

How could it be possible?

[Ram]

Does limma recommend you use percentages? How is a batch correction algorithm supposed to know the semantic limitations of your metric?

[Mary]

but limma takes normalized values, right? So my coverage was estimated in percentages, which states that it is a normalised value. That's why I used that as an input to limma

[jv]

the usage for removeBatchEffects states that the input matrix should be a

numeric matrix, or any data object that can be processed by getEAWP containing log-expression values for a series of samples. Rows correspond to probes and columns to samples.

Using percentages is not consistent with this expectation.

[Mary]

yes, as suggested https://support.bioconductor.org/p/9140296/#9156132 in this post, percentage values are explained how to deal with. However, my output values are in a range that I am NOT confident about. if u could just go through this, https://support.bioconductor.org/p/9140296/#9156132

Thanks