Fold change cutoff for downregulated genes

Question

[blz]

Hello everyone,

Sorry, it's such a basic question, but what's the meaning of a 1.5 fold change cutoff for downregulated genes?

For upregulated genes, in my understanding, it means a 50% increase. For example, considering the expression of the same gene in conditions A and B as

A = 100

B = 150

so,

B/A = 1.5

and I would say the gene is upregulated in condition B in relation to condition A, and B represents 150% of A or a 50% increase.

Is that correct? Or may I misunderstanding the meaning of fold change? And what's the equivalent logic for downregulated genes? I would appreciate any recommendation of material explaining these concepts (in the "for dummies" style ;)

Thank you!

Answer 1

[LChart]

Typically a log-fold change cutoff for significance would be in terms of absolute value, i.e., abs(logFC) > log2(1.5).

For down-regulation that corresponds to logFC < -0.584 or FC < 0.667 or roughly 2/3.

This makes sense, since if A = 100; B = 150 (in your example) then A/B = 2/3.

[blz]

This is the way I'm filtering the downregulated genes (logFC < -0.584 or FC < 0.667), but I don't really understand it...

In the example, FC = 0.667 means A represents about 67% of B, that is, a 33% decrease. But the opposite represents a 50% increase.

FC cutoff aims to filter for genes with the same degree of change in both directions (up and down). How FC = 0.667 and FC = 1.5 can be equivalent?

May I interpreting in a wrong way?

[LChart]

They are equivalent in a multiplicative sense as 1.5 x 0.667 = 1. You want to use a multiplicative sense in this case since you're measuring how "many times more" condition A is than condition B.

Another way to think of it is that if A is 150% of B, then B is 67% of A, so the equivalent "fold change for downregulated genes" corresponding 150% is therefore 67%.

[blz]

Maybe I can understand...

I can not explain in a general way, but I can think in terms of physical things like lengths of bars...

If you have a bar A with length 100 and a bar B with length 150, then the difference between them is 50 units.

You can say both bar A is 50 units smaller than bar B OR bar B is 50 units greater than bar A, but the difference is always the same, it doesn't matter the direction of comparison.

You can also say bar A has a 33% decrease in relation to bar B OR bar B has a 33% increase in relation to bar A (as the difference represents 1/3 of 150)??? Then, the degree of change would be the same???

Thinking this way, a FC cutoff of 1.5 would represent a 33% increase/decrease...

Answer 2

[ahmad mousavi]

Take a look at the following page :

Understanding up and down regulated genes

Understanding up and down regulated genes from LOG2 foldchange or foldchange

[blz]

Thank you for the link, but I've seen it before...

I think I can understand the log scale: for each unit you "walk" in log2FC, you double FC (that is, if log2FC = 1, then FC = 2; if log2FC = 2, then FC = 4; if log2FC = 3, then FC = 8; and so on) OR you get the half (log2FC = -1, FC = 0.5; log2FC = -2, FC = 0.25; log2FC = -3, FC = 0.125).

But I'm confuse about FC: what means FC = 1.5 when the expression of the gene is reduced (i.e. downregulated)? The word "fold" gives the ideia of "times", that is, something that is augmented, so I can not understand properly this concept for down genes...

[MB]

You explained it already quite well yourself: a negative LFC means that you have less expression = downregulation (because something that is negative on the logscale is division/fraction on the 'normal' scale); as you said, a LFC of -1 means a foldchange of 0.5 (i.e. only 50 %) compared to your reference condition. Keep in mind, that a (log)foldchange is the result of a comparison between two conditions like treatment vs. control. If you compare your treatment against a control and you find a gene downregulated, that also means that you will find it upregulated when you compare control vs. treatment. The sign gives you the direction according to the comparison that you asked for, and the number gives you how large the difference in expression is.

[blz]

Thank you for your answer, it also help me to formulate the thought above (specially the bold part "compared to your reference condition")!