-quantMode GeneCounts output big differences between stranded and total

[Magic_tm]

Hi all,

I am using STAR regularly but I just encountered a really weird output with the -quantMode GeneCounts option. 

For some of the genes the column 2 output is much much lower than the sum of the stranded columns 3 +4:

here is my output:

N_unmapped	598099	598099	598099
N_multimapping	382731	382731	382731
N_noFeature	112716	10608161	10694861
N_ambiguous	1179474	44749	44951
AT1G01010	273	133	140
AT1G01020	249	123	127

AT1G01030	34	18	17
AT1G01040	1174	706	718

...

AT1G32630	26	11470	11613
AT1G32640	0	11604	11453

As you can see for the vast majority of genes column 2 is roughly column 3+4. I know it's not exactly the sum but still. However, for some genes, like AT1G32640 its 11k for 3 and 11k for 4, and column 2 is 0? Is this an issue with the gtf, although I got it from the Arabidopsis website?

My command line:

$bin_DIR/STAR-2.6.1c/bin/Linux_x86_64/STAR \

--runThreadN 64 \

--readFilesCommand zcat \

--readFilesIn $Fastq_DIR/trimmed/"$Samples"_1.trim.gz $Fastq_DIR/trimmed/"$Samples"_2.trim.gz \

--genomeDir $Genome_DIR \

--outSAMstrandField intronMotif \

--outFilterType BySJout \

--outFilterIntronMotifs RemoveNoncanonical \

--quantMode TranscriptomeSAM GeneCounts \

--twopassMode Basic \

--outFileNamePrefix $Output_DIR/$Samples

I would be grateful for any tips you may have.

Cheers Thomas

[Alexander Dobin]

Hi @Magic_tm

if exons of these two genes overlap completely on *opposite* strands, the unstranded counting will yield 0 - all the reads will be considered "ambiguous".

The stranded counts, however, will be done towards the correspodning gene.

The easiest way to check this is to view annotations on a genome browser.

Cheers

Alex