Single-cell foundation modeling with species-nativeprotein tokens links regenerative competence across frog and mouse
Cang, H.; Sun, S.
Show abstract
Appendage regenerative capacity varies dramatically across species, developmental stages, and anatomical sites, yet comparing functional transitions across organisms remains difficult because gene vocabularies diverge. Cross-species single-cell analysis conventionally collapses divergent genomes to one-to-one orthologs--a reduction that is not neutral. Here, we construct a species-native input representation for allotetraploid Xenopus laevis that preserves duplicated L and S homeologs (96.39% feature coverage versus 53.14% under symbol collapse) within a frozen universal cell embedding (UCE). A causally validated tail-organizer contrast defines a portable vector competence ruler. While baseline representations (direct expression, SVD, Harmony) recover organizer identity, only species-native UCE preserves the stage-52-versus-stage-58 limb competence transition, which ortholog collapse reverses. Applied without refitting, the ruler distinguishes regenerative from fibrotic digit repair in adult mice and resolves an aligned component in state-balanced macrophages, an ordering reproduced by simpler representations. Preserving species-native gene vocabularies carries functional contrasts across evolutionary and genomic boundaries.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- NiCo Identifies Extrinsic Drivers of Cell State Modulation by Niche Covariation Analysis 96%
- Overloading And unpacKing (OAK) - droplet-based combinatorial indexing for ultra-high throughput single-cell multiomic profiling 95%
- Dictionary of human intestinal organoid responses to secreted niche factors at single cell resolution 95%
Similar papers in this journal
- Conserved epigenetic regulatory logic infers genes governing cell identity 95%
- Multiome Perturb-seq unlocks scalable discovery of integrated perturbation effects on the transcriptome and epigenome 95%
- Integrative, high-resolution analysis of single cell gene expression across experimental conditions with PARAFAC2-RISE 95%
Similar papers in this journal
- Multiplexing cortical brain organoids for the longitudinal dissection of developmental traits at single cell resolution 95%
- Reproducible single cell annotation of programs underlying T-cell subsets, activation states, and functions 95%
- Mapping single-cell developmental potential in health and disease with interpretable deep learning 95%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.