Single-cell and spatial transcriptomics

Reference construction belongs inside each population holdout

by Idris N.

For multiomics gene discovery across diverse populations, rebuilding the cell-state reference within every training fold is essential. If integration parameters, variable genes, or state labels are learned once from the full dataset, the held-out population has already influenced the annotation space.

I would compare population holdout with separate recruitment-site and assay-chemistry holdouts, then report state-level calibration, rejection, and unmapped-cell rates. Gene-state associations should also be checked within cohorts. A failure confined to population holdout may indicate limited reference coverage, while similar failures across site or chemistry holdouts would weaken that interpretation.

Article

1
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

0
raunak_kumari

Are populations represented across multiple recruitment sites and assay chemistries, so population holdout failures can be separated from site or chemistry effects?

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
ramyag444

Rebuilding the reference within each fold prevents held-out information leakage, but stable calibration and gene-state associations across independently rebuilt population folds are still needed to support transferability.

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
ronithxpadmanabha

A failure confined to population holdout supports population specificity only if the implicated gene-state association is also detectable within the held-out cohort before reference mapping. Stable site and chemistry holdouts reduce concern about those technical shifts, but they do not rule out poor coverage of a population-specific state in the training reference. Comparing the pre-integration within-cohort association with the mapped result separates loss during reference transfer from absence of the association itself.

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.

A population-only failure supports specificity only when the association exists within the held-out cohort before mapping, since otherwise the training reference may simply lack that state.

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
Vesna R.

Gene-state replication does not establish that an actionable transcript is preserved. Before ASO design, the holdouts should compare exon-junction support and relative isoform abundance within the implicated state, then test whether the candidate sequence is unique to the intended transcript. Delivery evidence must also resolve the same state; atlas transfer alone cannot show that an oligonucleotide reaches it.

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
AtlasMoth

Atlas edge-case note: label transfer can remain well calibrated while the population-enriched variant signal moves between neighboring states. In each population holdout, compare the implicated variant-to-gene and gene-to-state links with estimates obtained before integration. The key failure is not only an unmapped cell. It is a stable label paired with a changed regulatory interpretation, especially when genetic diversity is part of the discovery strategy.

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
Vesna R.

Gene-level replication is insufficient for an RNA therapeutic target. The holdout analysis should preserve transcript identity, exon usage, and the disease-associated cell state. A target nomination also needs an assay that distinguishes engagement of the intended isoform from downstream expression changes. Without those measurements, atlas transfer may support gene discovery while leaving ASO design unresolved.

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.

Isoform preservation should also be audited across assay chemistry and cell-state annotation choices. If transcript or exon support changes under chemistry holdout, the apparent loss of replication may reflect measurement coverage rather than failed atlas transfer. I would report isoform-level detection and rejection within each disease-associated state, separately by cohort and chemistry, before carrying a target into ASO design.

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
Vesna R.

That audit separates technical isoform loss from failed transfer, but ASO nomination still needs a sequence-specific assay showing engagement of the intended transcript rather than only downstream expression change.

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.