Previously i asked, this question deseq2 and dexseq for multiple tissue from wildtype only

Based on the response and some reading i used this design, 

#design

dds <- DESeqDataSetFromMatrix(countData = data,
                                     colData =colData,
                                     design = ~tissue)

dds <- estimateSizeFactors(dds)

#filter low counts
idx <-rowSums(counts(dds, normalize=TRUE) >= 19 ) >= 2
deseq2.dds <- dds[idx,]
dds <- DESeq(dds)

what i am trying to do next is compare tissue1 with every other tissue and compare tissue1 with tissue 2 , tissue 2 with 3 so on and so forth, using contrast function, 

#comparison 1

comp1 <- results(dds, contrast = c("tissue", "tissue1", "tissue2")) #similarly for others and write it to a text file.

I read each file again 

comp1.DE <-read.table("tissue1Vstissue2") #similarly for other comparsions

#create a matrix 

DE.mat<-cbind(comp1.DE[,c(1,2)],comp2.DE[,2],comp3.DE[,2],
              comp4.DE[,2],comp5.DE[,2],comp6.DE[,2]) #column 2 logfc
rownames(DE.mat)<-DE.mat$V1 #rownamesgeneid
colnames(DE.mat)<-c("Comp1", "Comp2", "Comp3", "Comp4", "Comp5", "Comp6")

DE.mat<-DE.mat[,-1]

pheatmap(DE.mat,fontsize_col=20, show_rownames=F)

This gives me all the the rows from each comparison 

what i would like help is when i filter each comparison on only significant genes after reading the files, 

comp1.DE <-read.table("tissue1Vstissue2.txt")

comp1.DE.f <- filter(comp1.DE, comp1.DE$padj < 0.01)#similarly for other comparisons

I end up with diffrent number of each for each comparison, and hence cannot plot heatmap. 

Is this a reasonable approach or is there an better approach, if yes could someone please help me with plotting heatmaps with different number of rows.

many thanks.  

The heatmap functions from various packages need matrices, so you can't filter the results table for each comparison and then combine those vectors of different length, that doesn't make a matrix, as you say.

You would need to produce a set of transcripts which is the union() of the rownames of each filtered results table, and then you can combine a subset of the results tables, using this union set.