Weird promoter ratio versus log10(UMI) curve

[strawberry789]

This is what my promoter ratio versus log10(UMI) curve looks like:

What's wrong with this? And how should I select the quality barcodes based on this graph? I tried to set low quality criteria, but result in zero cell barcodes after filtering.

Thank you.

[ChenManqi2]

For custom references or references missing the promoter.bed file, the count of promoter_region_fragments is 0. I use TSS_fragments to filter cell barcodes instead.

[murphchuyi]

My promoter ratio plots were exactly the same as yours 🌚, although my reference genome version is hg38, of which the bed file should not be missing, this is weird. I instead use the peak ratio combined with logUMI as QC metrics.