Zero FPKM values after stringtie quantification

Question

[Pedro]

Hi everyone,

I am carrying out RNAseq analyses in Arabidopsis using hisat2 as aligner and stringtie for merging and gene expression quantification (both versions 2.2.1). These are my instructions:

hisat2 --rna-strandness R --dta -x ../../genome/index -U sample.fastq.gz -S sample.sam

## Generating sorted bam file
samtools sort -o sample.bam sample.sam
rm sample.sam
samtools index sample.bam
bamCoverage -bs 10 --normalizeUsing CPM --bam sample.bam -o sample.bw

## Transcript assembly
stringtie -G ../../annotation/annotation.gtf -o sample.gtf -l sample sample.bam

## Preparing merge list file for transcriptome merging
  echo /home/pedro/arabidopsis/rnaseq_studies/co_rnaseq_2022/co_rnaseq_14dag_R/samples/sample/sample.gtf >> ../../results/merge_list.txt

## Gene Expression Quantification
stringtie -e -B -G ../../annotation/annotation.gtf -o sample.gtf sample.bam

My library type is single end and fr-firsstrand.

The point is that after stringtie quantification I find many genes which show a value of zero fpkms in one of the biological replicate and a non-zero value (high value) in the other of the replicate. For example, the gene AT1G43310 has 0, 0, and 11.445469 FPKM for the three biological replicates of one genotype. BUT, after looking the alignment in the IGV I find this is wrong.

Every sample analysis is performed in the same way and this happens for many genes. I have tried several options and looked for answers here but I am not able to solve it. Any suggestion is welcomed.

Thanks,

Pedro.

[R]

Dear Pedro,

Hello, Has this issue been resolved? I've encountered the same situation myself, so if it has been resolved, could you please let me know?

Answer 1

[i.sudbery]

Looking at the example in the image here, I would say that the expression should be zero in all three replciates. The expression shown isn't really compatible with the shown transcript in any of the repclicates, as there isn't any expression from exon 1.

In general, stringtie isn't recommended from quantitation. You should use stringtie only if you need to generate a sample specific transcriptome annotation. For a well annotated species, this shouldn't be neccessary for most simple A vs B gene-level expression based studies. If you do need to run stringtie, use it to generate the transcriptome annotation, and then quantify that annotation using something like Salmon or RSEM.