filter genes with no significant change in expression value

Question

[Zara]

Hi Everybody,

I'm trying to run a gene regulatory inference method. In the first step I need to remove genes with no specific change in their expression values. I don't know how to specify if a gene has "significant" change. What does it mean? My second question: Should I just remove these genes (not significantly changed) if I'm looking for their regulators? Or a gene which has not changed significantly can't be also detected as a transcription factor? In other words: I thought maybe a gene is not changed significantly, so I am not be able to find those genes regulating this gene. But what if I want to detect the influence of this gene on other genes? If it is not changed, is it not worthy to consider its effect on other genes? And gene regulatory network inference methods only rely on the changes to find regulatory interactions?

Thanks.

Answer 1

[WouterDeCoster]

You may want to use a variance filter: (in R)

data$vars = apply(data, 1, var) #Calculate variance across rows

data2 = data[data$vars >= quantile(data$vars, c(.80)), ] #Eliminate rows, here using 80% percentile

data2$vars <- NULL #Removing the variance column

Perhaps someone with more R experience can rewrite this to a more concise filtering :-)