missing gene counts

[Bernadeta D]

Hi Alex,

I'm looking at STARsolo cell-gene-matix files and it seems to me that some gene associate reads that I can observe in BAM file are dropped even from raw matrix file and I'm not sure why that would be. 

Attached is a screenshot for what I'm looking at: on IGV viewer I can see a read that has TTGTAGGTCCGAACGC cell barcode and is mapped to VTRNA1-3. I then look at the barcode and feature files from STARsolo output and find their row numbers (727833 and 17667 for cell and gene, respectively). Then I look at the matrix to find the rows associated with this particular barcode and gene, but though I find the cell barcode I cannot find the VTRNA1-3 read that, given that it is present in bam file, should be associated with it. I would understand if I'd be looking at filtered files and missing cells because they did not pass the filter but this is looking at the raw cell-matrix and I'm not sure why would these reads be dropped?

Thanks for any insights.

B.

[Alexander Dobin]

Hi Bernadeta,

it looks like only part of this read overlaps the gene - such reads are not counted towards the genes.

The present policy (following CellRanger) is that only reads that are contained entirely within exons of a gene are counted.

Cheers

Alex

[Bernadeta D]

Thanks Alex, this solved my confusion :) And would the use of --soloFeatures GeneFull change this behaviour?

[Alexander Dobin]

Yes, GeneFull will include all reads that overlap (>=1 base) the whole genic locus, i.e. it will include purely exonic reads, exon/intron reads, purely intronic reads, and exon/intergenic reads.

The latter are the ones that will be counted in your example.