Hello,

I'm not sure of the details of your problem exactly without a bit more information. Is the issue that the colours are not clear, or that the dendrogram linking the samples isn't clear? 

What program are you using to generate your heatmap? You can make quite a pretty heatmap in R using the ggplot package and the heatmap.2 function. There are a number of tutorials available online to walk you through this.

If the issue that you're running the heatmap colours globally, try normalising the data by either the row or column (depending on which the species are on). This can be done manually by dividing each value by the sum of its row or column (or if using R, try the {scale="row"} command). 

If the problem is related to the dendrogram, this post might be useful: https://stackoverflow.com/questions/21983162/how-to-expand-the-dendogram-in-heatmap-2

Hope this helps a bit!

I have not used heatmap,

By log transforming, one is bringing the values to a depictable range, but may not  give the desired graphical impression.

I have uploaded the file for you information. It has the list of counts and placed below are the heat maps. The first heat map represents the actual number, the 2 heat map is of log transformed. If you look at the log transformed its totally misinterpretating.

So is there a better way to represent it. Any suggestions

Hi Chetan,

From the looks of that, your heatmap is the data from the table flipped top-bottom, maybe try reversing the row order and see if it's any closer to what you want?

Ok I will give a description of what I need. I will put some species in this table which have a diverse range of count. Some may range between 0-100 and some species vary by 100-10000s. If I try to plot a heatmap with this values, it seems that the one which are low are of no importance or not shown up. I need to normalize this. I want to find a way where I can show for all species the same pattern i.e. low to higher, irrespective of their counts in various order. I hope Im not confusing you.

And also this is only a small part of the data. so there will be more species added in this table with some species representing at very low number but are ecologically important. So have to highlight this in that heatmap

Ah, I understand now. I think you need to normalise by row.

In a practical sense, what you want to do is add all the values in each row together, then divide each value by the sum. In example:

Initial

Species A: 1, 2, 3, 4, 0 (Sum of values = 10)

Species B: 100, 200, 300, 0, 400 (Sum of values = 1000)

Normalising:

Species A: 1/10, 2/10, 3/10, 4/10, 0/10

Species B: 100/1000, 200/1000, 300/1000, 400/1000, 0/1000

After Normalising:

Species A: 0.1, 0.2, 0.3, 0.4, 0

Species B: 0.1, 0.2, 0.3, 0.4, 0

Does that sound like what you're after?

Yes Andrew I am satisfied. I did what you suggested and it solved the problem, this can help a lot to answer my questions. Thank you all for the help.

Dear Andrew, can you say me the name of the normalization technique used in this thing. As i need to include it in the paper.

It's called feature scaling - we scaled to values between 0 and 1. Just say something along the lines of: "The data was scaled by the sum of each row"

Glad you got everything sorted okay!

Yes Andrew. Have a great time. Thank you.

Dear Andrew,

I would like to question you if this normalization also work to do the correlation of data which have a different scale ... I'm working with transcriptome and proteome data and I want to correlate the abundance values for common data but scales are different, example the max value by transcriptome is 16,01 and for proteome is 12,03 (after apply the log2 (count + 1) normalization for both datasets). The pearson correlation was very low which is arguable however I'm worry about the result being influenced by the differences in the scale and I would to find a way to correct it..... do you have some recomendation? thanks in advance....

https://reneshbedre.github.io/blog/hmap.html