Many reads assigned but with zero counts. Weird results

[biozhangzhou]

Hello,

I have three replicates of ONT cDNA sequencing files. A lot reads of them have assigned to a lnc gene, but the gene counts in one sample was zero. Why it happens ?

Thanks, Zhou

[andrewprzh]

Dear @biozhangzhou

Could you please check the assignment type of reads assigned to these genes? By default, only unique assignments are used in quantification. To change counting strategy you may use --transcript_quantification and --gene_quantification options.

Best Andrey

[biozhangzhou]

Dear Andrey: The sample with zero tpm value do have many unique reads (see the statistic of colume six for OUT.read_assignments.tsv below ). So it's really weird.

Best Zhou

[andrewprzh]

Dear @biozhangzhou

Indeed, looks odd. Could you send me just a small fragment of your BAM file (i.e. via email)? Privacy is guaranteed.

Best Andrey

[andrewprzh]

Dear @biozhangzhou

I now realized that if a transcript is spliced, it must have at least on spliced read assigned to it to deemed as expressed. Otherwise it is not reported in the count table even if it has a lot of unique, but non-spliced alignments. This check is done since it looks quite suspicious for a spliced transcript to have 0 spliced alignments. At the same time, mono-exonic alignments have a higher false-positive rate.

Could you please check if it's the case?

Best Andrey

[biozhangzhou]

Dear Andrey:

Very grateful for your continuous help and sorry for failing to respond. I have actually deleted that results and cannot reproduce that weird results now. Everything seems to be going well, but I will keep your advice in mind and give you feedback when I meet familiar problems.

Best Zhou

[andrewprzh]

@biozhangzhou

Thanks for getting back, it took me a while to realize.

Best Andrey