Heatmap: how to just show significantly differentially expressed transcripts in it ?

Question

[Farbod]

Dear Biostars, Hi

I have about 8000 transcripts that has hit with Transcription Factors (TFs). Among them 69 are up-regulated in females (3 biological replication) and 19 up-regulated in males (with 3 biological replication). (using DESeq2, FDR 0.05)

Now I intend to draw a heat map based on my TMM-normalized expression values (3 for males and 3 for females for each transcripts) but I want to just show the DEG in it (69+19) not all 8000 transcripts; something same as Figure8b in here.

1) How can I do it? is it just by selecting those 88 transcripts TMM value and used them as input in R heatmap script?

2) As you can see in the Figure8, there is just 2 columns. but as I have 3 BR for each conditions, I have 6 columns. How I can convert them into 2 columns ?

Can I calculate the mean of each 3 TMM of each transcripts and use the mean number as input for each transcripts ?

~ Thank you in advance

Answer 1

[TriS]

1) How can I do it? is it just by selecting those 88 transcripts TMM value and used them as input in R heatmap script?

create a matrix containing only the genes you are interested in and use that for your heatmap. you can easily do that in R (assuming genes in rows and samples in columns):

m <- myMatrixWithResults
index_genes <- which(rownames(m) %in% myListOfGenes)
sub_m <- m[index_m,]

2) As you can see in the Figure8, there is just 2 columns. but as I have 3 BR for each conditions, I have 6 columns. How I can convert them into 2 columns ?

I would prize you for showing all replicates to show that your results are consistent. I'd suggest to scale/center the data and use this matrix for heatmap. for the image itself you can use pheatmap, which is straightforward and offers pretty coloring options, including scaling the data.

[Farbod]

Dear @TriS, Hi and thanks for your help

1- My data are matrix of TMM-normalized expression values as produced by Trinity. Do I need to "scale/center the data" in this case ? if yes What exactly you mean by this?

2- I want ti use this color scheme. Do you know how to import it in the script of heatmap in R ?

3- I found this heatmap example very pretty but there is many error is running it.

~ Take Care

[TriS]

1- you can use pheatmap(myData,scale="row") 2- use color=colorRampPalette(c("#fee8c8", "#fdbb84", "#e34a33"))(100),breaks=seq(-2,2,length=100) within the pheatmap function 3- without your code it's impossible to say why you get the error

[Farbod]

Dear TriS,

Thank for all your helps and very useful guidance.

One problem that I have encountred is this that when I run my script (below), I could not see my 88 transcripts ID in my heat map (e.g TRINITY_DN107669_c1_g2_i1)

Would you please help me about it ? ( I even have tried my_matrix <- t(my_matrix) but no chance!)

this is head of head(my_data):

                samples    F1    F2    F3    M1    M2    M3

1 TRINITY_DN107669_c1_g2_i1 2.120 0.173 0.781 0.000 0.000 0.000

2 TRINITY_DN101742_c5_g1_i2 0.398 0.303 0.751 0.000 0.009 0.000

3 TRINITY_DN107731_c4_g1_i4 0.102 0.336 0.160 0.000 0.000 0.000

And this is my R script for Heatmap :

library(ComplexHeatmap)
library(pheatmap)
filename <- "male-female-88-TMM.matrix"
my_data <- read.table(filename, sep='\t', quote='', stringsAsFactors=FALSE, header=TRUE)
head(my_data)
dim(my_data)

row.names(my_data) <- d$sample
my_data <- my_data[, -1]
my_data <- my_data/rowSums(my_data)
my_matrix <- as.matrix(my_data) 

head(my_matrix)
samples_info <- data.frame(samples= my_data$samples)
pheatmap(my_matrix,scale="row", color=colorRampPalette(c("#fee8c8", "#fdbb84", "#e34a33"))(100),breaks=seq(-2,2,length=100))

[TriS]

what does head(my_matrix) return?

[Farbod]

> dim(my_data)

[1] 88 7

> head(my_matrix)

         F1         F2        F3         M1          M2          M3

[1,] 0.68965517 0.05627846 0.2540664 0.00000000 0.000000000 0.000000000

[2,] 0.27241615 0.20739220 0.5140315 0.00000000 0.006160164 0.000000000

[3,] 0.17056856 0.56187291 0.2675585 0.00000000 0.000000000 0.000000000

[TriS]

you don't see the sample names because you don't have rownames :)

[Farbod]

Dear TriS, Hi

I have corrected the rownames, finally. ;-)

actually it was in my script but as I have copied this script from different references, I have used d instead of my data in my script. so;

incorrect : row.names(my_data) <- d$sample

correct : row.names(my_data) <- my_data$sample

Now I have another label for rows; in addition to transcripts IDS (TRINITY bla bla bla), I have the gene names that showed in annotation and I want to add them in the different side of the rows (left side the IDs and right side the blast hits).

1) How can I do it ? did I must add another column to my original file containing the gene names or I can insert these lables from separate file ?

2) I have seen a shrink option in some heatmaps. have you any experience about using it in Heatmap() ?

3) in your first script (index_genes <- which(rownames(m) %in% myListOfGenes)) you kindly have shared with me, what did %in% myListOfGenes mean?

[TriS]

1. I personally don't know a way to add a second column with the different rownames as text.
2. never used it
3. the "%in%" command allows you to look for something into something else, and returns the index with the position(s) of the element(s) in the array you searched. in that case I assumed you had a list of gene names in myListOfGenes and looked at where those were rownames of the matrix m...this is a pretty basic R command, I would suggest you to check out some online tutorials and play a little with value matching

[Farbod]

Hi again,

I need to show my 88 row names at the side that I am showing my clustering/dendrogram (left side).

They are now on the opposite sides.

in Heatmap(), it could be done using row_names_side = "left"androw_hclust_side = "left",BUT I want its option in heatmap() and pheatmap().

Do you have any suggestion dear TriS ?

[TriS]

hi Farbod, I am not sure how to do that, it might be useful to start a new thread with that question since it's drifting away from your initial problem