Dear  Michael Love:

I used fpkm(dds) form DESeq2 R package to calculate fpkm , the progress did not show any error.

The I summed all gene's fpkm value belong to 1 sample, and the sum is 400003.3, which is far from 1000000, I just want to check if i made some mistake during the calculation.

Here is the comand line I used ：

raw_count<-read.csv(file.choose(),header = T,row.names = 1)# input raw_count
head(raw_count) # check

condition <- factor(c('A','A','A','B','B','B'))# treatment

dds <- DESeqDataSetFromMatrix(raw_count, DataFrame(condition), ~ condition)# dds
nrow(dds)# check

G_range<-read.csv(file.choose(),header = T,row.names = 1)# Grangelist were obtained form gff3 file, just like the table showed below.

G_range_list<-makeGRangesFromDataFrame(G_range)# get list
rowRanges(dds) <- G_range_list

#Cal fpkm
fkpm_deseq2_dds<-DESeq(dds)
fpkm(fkpm_deseq2_dds)
head(fpkm(dds))# check 

write.csv(fpkm(fkpm_deseq2_dds),'G:/fpkm.csv')#output FPKM

FPKM don’t sum to 1,000,000, the standard TPM and CPM/FPM do. But also we have a robust version of normalization such that even our “FPM” don’t sum to 1,000,000 (it is controlled by an argument which you can turn off, see the help). This makes them robust for comparison across samples for cases such as: one highly expressed gene changes its level across sample but none others do. Robust normalization can handle this while strict compositional measures will induce changes to all genes. You may want one or the other so we have an argument that controls it.