Comparing gene expression between two structures using log2(intensity) values

Science
, , , ,
1

Hi all!

I am new to this forum and to gene expression analysis, being this the first time I use data from the Allen Human Brain Atlas. I am trying to compare the expression of a particular gene (e.g. GRIN1) between two structures (e.g. cerebellar cortex [CbCx] and cerebellar nuclei [CbN]). For this I did the following:

  1. I used the Microarray data from the Allen Human Brain Atlas. I did a “Gene search” since with the “Differential search” I could not easily find a gene among all the thousands included in that search.

  2. I set the Heatmap resolution to “Coarse” and sorted data by structure instead of by donor.

  3. Following what I read in this Topic and in this other Topic I changed Color map to log2(intensity).

  4. I downloaded the data and worked on Excel. As recommended there, I choose the probe with the highest log2(intensity) values/warmest colors (i.e. probe A_23_P84670).

  5. Then, as it was also recommended there, I averaged the log2(intensity) values from all 6 donors for each one of the two structures I am interested, obtaining these results: average(CbCx)=11.0799666 and average(CbN)=11.76543333

  6. I calculated the difference between the averages, which, if I understand well, would give me the fold-change in gene expression between the structures, which gave me a negative value:
    log2(FC) = average(CbCx)-average(CbN) = 11.0799666 - 11.76543333 = -0.68547

  7. Then I calculated FC as follows, obtaining a value below 1:
    FC = 2^(log2(FC)) = 2^-0.68547 = 0.62180466

Following the same steps but for another gene (GRIN2C), after calculating the difference between the averages I obtained a positive log2(FC) value:

log2(FC) = average(CbCx)-average(CbN) = 13.34336667 - 10.68195 = 2.661416667

So FC was above 1:
FC = 2^(log2(FC)) = 2^2.661416667 = 6.326539845

My doubt is what exactly this FC values (if calculated correctly) would mean, i.e. how should they be interpreted if they are below or above 1. If, for example, I take the same structure there will be no actual change in gene expression. For example, in the case of GRIN1:
log2(FC) = average(CbCx)-average(CbCx) = 13.34336667 - 13.34336667 = 0
FC = 2^(log2(FC)) = 2^0 = 1
Would a FC of 1 mean that the expression level is the same?

Thank you very much for your help!

2

Hi @Carlos,

Your analysis makes sense and should give you the result you want. As for what FC (or “fold change”) means, it a quantitative metric of how much more (or less) expression a gene has in one brain structure relative to another. So in the example:
log2(FC) = average(CbCx)-average(CbN) = 2.66; FC = 6.32
This means that expression of GRIN2C is 6.32 times higher in CbCx than in CbN.

A negative value would indicate the opposite, as in this example:
log2(FC) = average(CbCx)-average(CbN) = -0.68; FC=0.62
This means that expression of GRIN1 is higher in CbN than CbCx (e.g., 0.62 times as high in CbCx than in CbN).

You can get the direction immediately from the average values in your step #5 as well.

Finally, a value of FC=1 means no change. There is never a change in expression when comparing a value to itself. Hopefully, this all makes sense.

3

Thank you very much for your answer, @jeremyinseattle ! I guess that in my newcomer mind I was somewhat confused by the fact that FC=1 means “no change” (letting apart the calculations) since it would seem to me that “no change” would be the same as “no fold-change” and, thus, “FC=0” (again, letting apart the calculations).

Once again, thank you very much!

4

Happy to help. This is why we typically report log2(FC) because 0 = no change :slight_smile:

1 Like
5

Now it all makes sense! Thanks a lot!