Hi, We have tried several samples from an experiment in GEO (Series GSE74411), especifically SRA accessions SRR2833398, SRR2833399 and SRR2833400; and in all of them we have a larger number of reads alignments to forward (>90%) than to reverse (<10%) strands. The experiment says it uses a paired-end library and a stranded construction protocol, according to GEO.
We made a pretty straightforward analysis:
hisat2-build Spombe.fasta SPindex hisat2 -p 8 -x SPindex --sra-acc SRR2833398 -S SRR2833398.sam samtools sort -@ 8 -o SRR2833398.bam SRR2833398.sam samtools index SRR2833398.bam bamCoverage -b SRR2833398.bam -o SRR2833398_f.bw -p 8 -bs 1 --normalizeUsingRPKM --filterRNAstrand forward bamCoverage -b SRR2833398.bam -o SRR2833398_r.bw -p 8 -bs 1 --normalizeUsingRPKM --filterRNAstrand reverse
The result is therefore a wig with 10x coverage levels on forward than reverse strands. We can of course normalize, but is it normal to have much larger coverages in the forward strand? If so, why? We think it should be more or less the same according to the technique. If not normal, any ideas what can we be doing wrong?