Question

Conversion of LogFC to FC

2

Entering edit mode

nia ▴ 30

@nia-12707

Last seen 4.7 years ago

Hi everyone, I have write a script to analyze microarray illumina data by using Limma package. The script generated a table and it consist a logFC column. Now, I wish to know the FC values but here I have a confusion that should I go with 2^logFC or 10^logFC. I am looking for help.

limma logfc fc • 25k views

ADD COMMENT • link updated 4.7 years ago by Gordon Smyth 51k • written 4.7 years ago by nia ▴ 30

score 3 · Accepted Answer · 2020-03-07

3

Entering edit mode

Gordon Smyth 51k

@gordon-smyth

Last seen 3 hours ago

WEHI, Melbourne, Australia

Just type ?topTable to see that logFC is log2 rather than log10.

ADD COMMENT • link 4.7 years ago Gordon Smyth 51k

0

Entering edit mode

Thankyou for the quick response, that means I have to go with 2^logFC.

ADD REPLY • link 4.7 years ago nia ▴ 30

1

Entering edit mode

Yes.

Alternatively, you can use sign(logFC)*2^abs(logFC) which has some symmetrical properties.

For a logFC of -3, instead of getting 2^-3 = 1/8 = 0.125, you get -8.

Here you have to interpret -x as 1/x. This could help. Of course, there is no in ]-1:1[.

> 2^3
[1] 8
> 2^(-3)
[1] 0.125
> sign(-3)*2^abs(-3)
[1] -8
> sign(3)*2^abs(3)
[1] 8

ADD REPLY • link 4.7 years ago SamGG ▴ 350

0

Entering edit mode

Very detailed information thank you. The point I learn is if I do 2^logFC then I have to interpret that -logfc as 1/-logfc.

ADD REPLY • link 4.7 years ago nia ▴ 30

0

Entering edit mode

No I understand what you mean. Here is what I mean:

If logFC = 3, then FC = 2^3 = 8

If logFC = -3, then FC = 2^-3 = 1/8 that can also be noted -8 using the convention that -x means 1/x

ADD REPLY • link 4.7 years ago SamGG ▴ 350

1

Entering edit mode

I can't help but feel that this notation causes more problems than it solves; having the magnitude of the fold-change and the sign of the log-fold change is begging to be mistaken for one or the other. Certainly, if I saw a negative "fold change" in an analysis report, I would assume that the author made a typo and actually meant "log-fold change".

If one finds log-values or fold-changes below 1 too difficult to comprehend, a more mathematically appropriate notation would be to state 1/x rather than -x, where x = 2^abs(logFC) for logFC < 0. Then you get the magnitude of the fold change without repurposing the minus symbol to represent something other than its intended meaning.

ADD REPLY • link 4.7 years ago Aaron Lun ★ 28k