Article

Mechanistic Interpretability of Biological Foundation Models: An Empirical Analysis of scGPT Gene-Embedding Geometry

2026-07-24

Abstract excerpt

<title>Abstract</title> <p>Biological foundation models are often claimed to learn structured biological knowledge, but many interpretability studies still rely on qualitative examples or downstream task scores. This paper asks a narrow empirical question: does the learned gene-token embedding space of scGPT encode known physical interaction structure among human gene products? We analyzed the public scGPT human...

Topics

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

Identifiers and source

Literature Corpus work: eb2a7fc4-9b64-5a40-9b99-d4797b7eaf2a
DOI: 10.21203/rs.3.rs-10461111/v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Mechanistic Interpretability of Biological Foundation Models: An Empirical Analysis of scGPT Gene-Embedding GeometryDOI 10.21203/rs.3.rs-10461111/v1Read

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

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

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

Geometric Representations of Knowledge Inside Biological Large Language Models: an Empirical AnalysisSemantic proximity Read

Building Foundation Models to Characterize Cellular Interactions via Geometric Self-Supervised Learning on Spatial GenomicsSemantic proximity Read

scGenePT: Is language all you need for modeling single-cell perturbations?Semantic proximity Read

Domain-specific embeddings uncover latent genetics knowledgeSemantic proximity Read

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

A Gene Set Foundation Model Pre-Trained on a Massive Collection of Diverse Gene SetsSemantic proximity Read

Benchmarking gene embeddings from sequence, expression, network, and text models for functional prediction tasksSemantic proximity Read

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

CLOP-DiT: Structured-Metadata-Conditioned Single-Cell Latent Generation via Contrastive Language-Omics Pretraining and Diffusion TransformersSemantic proximity Read