Skip to content
Open access

Benchmarking single-cell foundation models for aging biology

Sep 2026 · bioRxiv · 0 citations · 52 references
Biology

TL;DR

Overall, scFMs supported diverse aging analyses, but performance depended on the biological question, highlighting their utility for rare cellular state identification and regulatory analysis.

Abstract

Single-cell foundation models (scFMs) provide representations of cellular states, but their utility across biological questions in aging research remains unclear. We established a benchmark of cellular representations for aging research, evaluating ten general-purpose scFMs, three aging-specific models and conventional methods across five biological questions using more than2.5 million single-cell transcriptomes. Using frozen pretrained representations, Geneformer performed best among scFMs for chronological-age prediction and age–pseudotime concordance, although 2,000 highly variable genes achieved higher mean performance. Several scFMs captured positive molecular-age shifts across three disease contexts, consistent with reported aging-associated changes. SCimilarity performed well for rare cellular state identification across out-of-distribution datasets, exceeding aging-specific models and conventional baselines. At the gene level, scGPT showed the highest recovery of reference TF–target interactions, including aging-related regulatory hubs. Overall, scFMs supported diverse aging analyses, but performance depended on the biological question, highlighting their utility for rare cellular state identification and regulatory analysis.

Read PDF

Similar papers

Open access Sep 2026

A single-cell view of human tissue aging reveals architectural decline beyond cellular composition

Treating the GTEx histopathology archive as a population-scale, life-span-resolved resource for spatial biology, single cells were detected across 16 human organs from nearly one thousand individuals and architectural remodeling was largely decoupled from cell composition, showing that human tissues age along two partl...

Ernesto Abila, Yimin Zheng, Z. Bagó-Horváth et al. · 0 citations
Sep 2026

Tissue-specific aging clocks map structured aging-modulatory drug-score patterns across 49 human tissue and cell-line categories

These findings describe category-associated drug-score patterns and provide a tissue-aware resource for generating hypotheses about drug responses, while cautioning against over-interpretation of individual drug-category predictions without experimental validation.

Dong-Dong Zhang · 0 citations
Review Open access Sep 2026

Tissue-specific aging—a new paradigm based on transcriptional insights from select model systems and humans

Aging is a complex process affected by genes and modified by environment. The forefront of aging research has moved from the level of organisms to the level of tissues and single cells, analyzed using bulk tissue and single-cell transcriptomes. Transcriptomics provides high resolution, comprehensive information per sam...

Stewart A. Frankel, B. Rogina · 0 citations
Open access Oct 2026

Beyond the Final Layer: Evaluating Single-Cell Foundation Model Representations for Molecular Prediction

Geneformer, scBERT, scGPT, and scFoundation are evaluated as frozen feature extractors for cell-type and perturbation-response prediction using layer-wise probes, leave-one-dataset-out (LODO) selection, readout controls, representation diagnostics, centered kernel alignment (CKA), and retrieval/collapse analyses.

Ming Wen, Yiwei Lyu, A. Robbins · 0 citations
Review Open access Oct 2026

Modeling human neurodegeneration with iPSCs: opportunities, limitations, and the challenge of aging

A framework matching model systems to biological questions is proposed, improving mechanistic understanding and therapeutic development and addressing the aging problem: reprogramming erases epigenetic signatures of donor aging, removing the cellular context in which late-onset sporadic neurodegeneration emerges.

Shi-Jie Wang, L. Elia, Wei-Yi Liu et al. · 0 citations
Open access Sep 2026

scTransMIL bridges patient-level disease states and single-cell transcriptomics for cancer screening and heterogeneity inference

It is demonstrated that scTransMIL accurately predicts sample-level cancer phenotypes, including the tissue-of-origin for metastatic tumors, and robustly identifies tumor-associated cell populations and maps biological progression trajectories using only minimal sample-level annotations.

Zhen-Chao Tang, Fang Wang, Fan Yang et al. · 2 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.