Very low % of reads showing primary alignment to transcriptome

[Asmita]

Dear users,

I have recently started analyzing RNA-Seq data for gene expression analysis, hence am quite new to the field.

I have used STAR for aligning RNA-Seq reads (hg38, Ensemble, release 94) using --quantMode TranscriptomeBAM for STAR run.

When I analyzed the quality of BAM files (2 files - genomic BAM and Aligned.toTranscriptome.bam) using BamQC, I get widely different results in terms of basic statistics like primary alignments

In whole in genomic BAM, total 96.95 reads fall in primary alignment, transcriptome BAM has only 29.4% primary aligned reads. Does this low % means the data quality is bad for doing analysis like differential isoform and allele expression?

Thanks for your inputs.

[GenoMax]

Have you tried to see where (96-95-29.4) reads are aligning (since they are not aligning to transcripts)? Does your data have rRNA present? Inspecting the resulting BAM using IGV would be a great place to start.

[Asmita]

Thanks for the response!

I did try to visualize the two BAM files in IGV.

When visualizing genomic BAM (Aligned.out.bam), I can see that many reads are falling into the exonic region of genes, with corresponding higher coverage, however, the coverage is missing in transcriptomics BAM file for the same region.

I would expect the coverage from transcriptomics BAM file to exist at least in genes whose annotation is present in the GTF files used for mapping. (Here the visualization is over part of MYH9, myosin 9 gene, ENSG00000100345)

As an additional note, when I analyzed my BAM (genomic) file using PICARD tools, I observed that 45% of total input bases were classified as intronic bases, while 50% of total bases were categorized as mRNA bases.

Is that the reason why there is low %reads in transcriptome BAM?