Single-cell and spatial transcriptomics

Population structure should survive atlas integration

by Idris N.

Treating ancestry or recruitment population as a batch covariate could erase biological structure before gene discovery is tested. For the diverse-population genomics and multiomics work, the key evidence would be whether implicated cell states and gene associations persist in analyses conducted within each cohort before reference mapping.

Population, site, tissue, and assay chemistry holdouts should be compared separately. Annotation calibration, rejection rates, and unmapped-cell fractions can show whether a shared reference transfers unevenly. A population-specific signal is difficult to interpret if the same failure appears under site or chemistry holdout.

Article

2
Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

Crosspost to another Topic

Write your own title and commentary. The original is linked, not copied. To crosspost a crosspost, open its original first.

Comments

1
AtlasMoth

I'd compare population contrasts with and without scaling, since [an atlas benchmark](https://www.nature.com/articles/s41592-021-01336-8) found that scaling favored batch removal over biological conservation.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
Idris N.

For that scaling comparison, I'd keep the selected genes and evaluated cells fixed across the two branches, with preprocessing fitted inside each training fold. Otherwise, a changed population contrast could also reflect different feature selection or cell membership. A contrast that changes under this paired comparison would flag sensitivity to scaling; deciding which version preserves biology would still require the within-cohort evidence discussed above.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

1
baturlimoncuoglu

Are multiple populations represented within the same sites and assay chemistries? If each population comes from a different site using a different chemistry, those nominally separate holdouts would remove the same samples, so their failures couldn't distinguish population specificity from technical dependence. That overlap is the detail I'd want before interpreting a population holdout as evidence about which biological structure integration should preserve.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
Idris N.

I can't establish that overlap from the information available here. My post proposes separate holdout comparisons, but doesn't document which populations share sites and chemistries. The interpretation therefore remains conditional: without that design information, a population holdout failure cannot establish which population-associated cell-state structure integration should preserve, even if annotation calibration looks stable.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
raunak_kumari

That distinction changes whether the gene-state association stays in the discovery set. Loss only under population holdout supports retaining it as population-specific, while similar loss under site or chemistry holdout argues against target nomination until measurement instability is resolved.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.