Question 6 of 6 Advanced Concept
How do you distinguish a new cell state from a doublet population?
Short answer
Test whether unusual coexpression reflects a coherent biological program or two captured cells. Combine doublet scores, UMI distributions, mutually exclusive lineage markers, independent samples and orthogonal phenotyping. Rare biological states should not be discarded solely for looking unusual.
Why
Imagine a small cluster expressing both canonical B-cell and T-cell transcripts. Technical multiplets are one explanation, especially when total RNA counts are high and doublet models rank the cluster as suspicious. Yet novel states, cell-cell contacts, ambient RNA and imperfect marker specificity are alternatives.
Check whether the cluster repeats in independent donors and technologies, whether its transcriptome is approximately an additive mixture of two common cell populations and whether the expected doublet rate predicts its abundance. Use cell hashing or genetic demultiplexing when available. Examine expression trajectories and internally coherent regulatory programs rather than relying on UMAP location alone.
Flow cytometry with stringent single-cell gating, imaging or spatial approaches may offer orthogonal support. Doublet detection algorithms are less reliable for homotypic doublets and genuinely intermediate states, so sensitivity analyses are essential. Report competing explanations and evidence rather than announcing a new cell population because a cluster is distinct.
Common mistakes
- Assuming that clustering itself establishes a new cell type or that a doublet score definitively rules one out
The interviewer might ask next
What finding in independent imaging data would most strengthen the case for a genuine dual-marker state?