I'm not really sure it helps with your original question. It's not really a coverage metric of any sort, but more of a way to rule in a "real" detection event instead of a false positive.

I've spent the last week working to decide how to adopt/adapt/extend Tax Triage. The use case for that metric is just to capture the intuition that one poorly mapped read repeated 100x is not the same as 10 decently mapped reads scattered across the same genome.

The machine hummed, warm and constant, like a partner with nothing to say. It was a lonely racket, genome assembly. Nothing but you, the code, and a pathogen nobody wanted to meet. (4/4)

I leaned back, eyes gritty from hours spent squinting at the screen, watching the raw data flicker by like the faces of strangers in a crowd. They were fragments, pieces of something bigger, something hidden. (3/4)

Me? I was elbows-deep in the guts of a bacterial genome, trying to make sense of what I’d scraped together from a handful of low-quality reads and the vague hope I’d get a hit. (2/4)