How to booleanize gene expression data from 1 cell type.

Question

[Bioinformatist Newbie]

Usually we do differential analysis between two states of interest (control vs. disease) or two cell types (MCF7 vs. HL60) to find gene which are up and down regulated 'differentially'.

What we should do when we are interested in knowing genes which are up and down regulated in only 1 cell line of interest. How to define the threshold score for gene being expressed and not expressed?

Answer 1

[Devon Ryan]

That's not "up or down regulated", that's expressed or not. The answer is that if you get any uniquely aligned reads and you DNase treated the samples and your reference is high quality then the gene is expressed. Whether it's meaningfully expressed is another question, which would be answered by looking at the TPM distribution.

[Bioinformatist Newbie]

I didn't exactly get your answer. Suppose I have a 'transcription profiling by array assay of GM12878'. I want to find out what genes are expressed/not-expressed in it, how I can find out this?

[Devon Ryan]

In the specific case of arrays, some of them have negative control probes, against which you can compare things.

[Bioinformatist Newbie]

You mean which have positive expression values they are expressed and which are having negative value they are not expressed. Right?

[Devon Ryan]

Google "negative control".

[TriS]

if a gene not expressed it won't be detected. a negative value generally means down-regulation in respect to a control sample and it's expressed as log2(FC). within a single sample, a negative value could simply be the normalized probe intensity value, but it does not mean that your gene is not expressed. generally speaking, if you see a value it means the gene has been detected and the higher/lower the value, the higher/lower the expression of the gene is...but still, if the gene is not expressed it won't be detected.