Is that a cell or an empty drop?

The problem. In droplet single-cell (day 9’s 10x method), most droplets contain no cell — just ambient RNA floating in the suspension. These empty drops still capture stray transcripts and get barcodes, so the raw data mixes real cells with thousands of low-count noise barcodes. A naive count threshold either discards small real cells or admits empty ones.

The idea. EmptyDrops models the ambient RNA profile explicitly: it estimates the background expression pattern from the clearly-empty droplets, then tests each barcode for whether its expression deviates from that ambient profile — not merely whether it has high counts. Small but genuine cells that differ from background are retained; high-count barcodes that just reflect ambient RNA are flagged.

Why it matters. This is the QC consequence of understanding the assay. Because I read how droplets generate the data, the need for this step is obvious rather than mysterious — ambient contamination is baked into the method. It’s the same lesson as fastp and MultiQC: knowing where artefacts come from tells you which cleaning step to trust. Scanpy and Seurat both lean on this logic.

Verdict. Foundational for single-cell QC and a standard early step. Read it for the statistical move — test against the ambient profile, not a raw threshold — and as a model of QC that flows directly from assay mechanics.