Hi,

I have total RNA-seq data and would like to compute TPM and FPKM values for a read count matrix generated for lncRNAs using the annotation file gencode.v28.long_noncoding_RNAs.gtf. To compute TPM and FPKM from the counts I first need the lengths of the lncRNA genes, which I obtain by summing their exon lengths accounting for overlaps using the R package GenomicFeatures. Then I compute effective length as gene length minus average fragment length + 1 (average fragment length was average insert length for any sample as provided to me by the sequencing lab). I used the following code:

library(GenomicFeatures)
txdb <- makeTxDbFromGFF("/home/inah/MESA_mRNAseq_New/gencode.v28.long_noncoding_RNAs.gtf", format="gtf",organism="Homo sapiens")
exons.list.per.gene <- exonsBy(txdb,by="gene")
exonic.gene.sizes <- lapply(exons.list.per.gene,function(x){sum(width(reduce(x)))})
exonic.gene.lengths <- as.numeric(exonic.gene.sizes)
names(exonic.gene.lengths) <- names(exonic.gene.sizes)
summary(exonic.gene.lengths)
#   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
#     68     499     734    1440    1775  205012

35 of the 15779 exonic.gene.lengths values are less than the average fragment length of 134 (for one sample).  So something must be going wrong here. So the effective gene length would be negative for these 35 lncRNAs. 

Does any GenomicFeatures (or other) expert have an idea of what is going on here? Am I calculating the lengths correctly? Is this maybe a problem with lncRNAs?

Thank you ... Ina