Article

Do Geometric Outliers Identify Important Genes in Single-Cell Foundation Models?

2026-06-27

Abstract excerpt

Single-cell foundation models (scFMs) produce gene embeddings that are increasingly interpreted biologically, but it is unclear whether different models organise gene space consistently, or whether geometric extremes identify important genes. The same four-metric screen was applied to Geneformer, scGPT and scFoundation, which differ in architecture, training objective and expression encoding. The models agreed wea...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 09287ac6-770b-5197-b3b9-c810797bf258
DOI: 10.64898/2026.06.22.733850

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Do Geometric Outliers Identify Important Genes in Single-Cell Foundation Models?DOI 10.64898/2026.06.22.733850Read

scRepresenter: a workflow for computing, integrating and benchmarking cellular representations in single-cell transcriptomicsSemantic proximity Read

Mechanistic Interpretability of Biological Foundation Models: An Empirical Analysis of scGPT Gene-Embedding GeometrySemantic proximity Read

When batch correction corrupts gene expression: uncovering distortions in correlation structuresSemantic proximity Read

<i>Milo:</i> differential abundance testing on single-cell data using k-NN graphsSemantic proximity Read

Coding agents author interpretable single-cell embedding models from the literatureSemantic proximity Read

Scoring gene importance by interpreting single-cell foundation modelsSemantic proximity Read

Reading Mechanism off Biological Foundation Models: An Empirical Analysis of Genomic Neighborhood Structure in scGPT Gene EmbeddingsSemantic proximity Read

Do Single-Cell Models Learn Real Biology? An Empirical Analysis of Physical Interaction Signal in Pretrained scGPT Gene EmbeddingsSemantic proximity Read

Sensitive cluster-free differential expression testing.Semantic proximity Read

Learning Minimal Gene Programs for Disease-Aligned RepresentationsSemantic proximity Read

PertEval-scFM: Benchmarking Single-Cell Foundation Models for Perturbation Effect PredictionSemantic proximity Read

Cell-attribute aware community detection improves differential abundance testing from single-cell RNA-Seq dataSemantic proximity Read