How to include genes having negative fold change in the log2 FC plot?

Question

[Susmita Mandal]

Hello everyone,

I am doing expression analysis on samples having different cell types and their conversions. While doing the fold change of few genes w.r.t. to one cell type, I am getting some negative values of few genes in the other cell types and thus I cannot include them in the log2 FC plot. I would like to incorporate those genes into the plot? Is there any way or I have to remove them and make the plot (sadly)?

Thanks in advance, Susmita

Answer 1

[swbarnes2]

If the fold changes are negative, they've already been logged.

[ATpoint]

Explanation: Imagine two values, x and y, and you divide them by each other (x/y) there are two options if values are not logged:

1. x >= y will give you results from 1 to infinity.
2. x < y will give you a result between 0 and 1

Obviously compressing values between 0 and 1 is not optimal. This is evident when you plot values from option 1 and 2, which will result in a highly assymetrical graph (because the space from 1 to Inf is visually larger than between 0 and 1). Therefore, one uses log2(x/y) which will transform fold changes from option 2 into a range of equal size as option 1.

1000 / 10 = 100
10/1000 = 0.01
=> pretty wide range



but:
log2(1000/10) = 6.64
log2(10/1000 = -6.64 
=> symmetrical around 0

You should note that no negative values can be calculated from non-logged count data., as any x < y will be somewhere between 0 and 1.