I am trying to subset a df of normalized counts and then make a heatmap using pheatmap() (essentially subsetting the heatmaps), but am running into the NA/NaN/Inf error. Interestingly, I am successfully able to create a heatmap with the non-subsetted normalized counts, but once I subset it, I am unable to generate the heatmap. Below is a snippet of my code. Thanks!

Also, I subsetted the df_normalized_counts_2 (rowxcol = genexsample) by the genes and didn't run into this issue (code for it not shown)

vsd_2 <- vst(dds_2, blind=FALSE) # generating a matrix of values for which variance is constant across range of mean values
df_normalized_counts_2 <- assay(vsd_2)

df_subset_3 <- df_pre_subset[df_pre_subset$condition_and_cellline %in% comparison_3,, drop = FALSE] 
subsetted_samples_3 <- rownames(df_subset_3)
subsetted_samples_3

df_normalized_counts_3 <- assay(vsd_2)[,subsetted_samples_3]


options(repr.plot.width=7, repr.plot.height=5)

pheatmap(df_normalized_counts_3.2,
         cluster_rows = TRUE,    # Cluster rows (genes)
         cluster_cols = TRUE,    # Cluster columns (samples)
         scale = "row",          # Scale rows (genes) for better visualization
         color = colorRampPalette(c("green", "black", "red"))(100),  # Choose a color palette
         main = "DEGs Heatmap",  # Heatmap title
         fontsize = 8)     # Adjust fontsize for gene names

It would something like this:

Heatmap non-subsetted:

Heatmap subsetted:

Here the error probably means that you habe genes where expression scross all samples is the same. Rowscaling is value minus mean divided by standard deviation. If standard deviation is zero then you get an infinite value which is what the error tells you. Subset to rows where not all values are the same.

Generally why scaling is beneficial, seehttps://www.biostars.org/p/466473/#466476