MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation

Abstract

Modeling gene regulation in single cells requires jointly representing DNA sequence, chromatin accessibility, and gene expression across regulatory scales. Sequence-based models capture DNA-level regulatory grammar but often lack cell-level resolution, whereas many single-cell omics models represent genes and peaks as fixed-vocabulary tokens without their underlying sequence. Here, we present MUSS, a sequence-based multi-scale model for single-cell gene regulation that preserves both DNA-level and cell-level resolution. MUSS derives peak- and gene-level representations from DNA using a frozen pre-trained sequence encoder and integrates them through peak-peak self-attention and distance-biased peak-gene interactions. Peak masked language modeling followed by training on paired scATAC-seq and scRNA-seq aligns these representations with cell-specific regulatory states. Across gene expression prediction, enhancer-gene linking, motif-guided regulatory prioritization, and zero-shot cell clustering, MUSS achieves strong performance against sequence-based and single-cell omics baselines. It further predicts expression for genes and peaks unseen during task training. These results support sequence-grounded, multi-scale representations for decoding cell-type-specific regulatory programs. Code, datasets, and model checkpoints are available at https://github.com/Lotfollahi-lab/MUSS_neurips26.

Population-scale single-nucleus multi-omic profiling of skeletal muscle reveals extensive context-specific genetic regulation
Arushi Varshney, Nandini Manickam, Peter Orchard +33
Nature Genetics28 Oct 2026Peer-reviewedLicence not stated22 citedMolecular biology & gene regulation Genetics & genomicsOpenAlex
A context-aware temporal modeling through unified multi-scale temporal encoding and hierarchical sequence learning for single-channel EEG sleep staging
Amir Ali Vakili, Salar Jahanshiri, Armin Salimi-Badr
PLOS One30 Sep 2026Peer-reviewedCC BY 4.0Full textNeuroscience PhysiologyPLOS
Codon-aware multi-scale feature fusion for metagenomic sequence classification
Yan Qian, Quanjie Song, Xinyi Xie +1
PLOS One5 Oct 2026Peer-reviewedCC BY 4.0Full textBioinformatics & computational biology Genetics & genomicsPLOS
Multi-model biological and sequence information fusion for gene regulatory network inference from single-cell transcriptomics
l. zhong, B. Yan, J. Wang +1
bioRxiv13 Sep 2026PreprintCC BYFull textBioinformatics & computational biologybioRxiv
A single-nucleus multi-omic atlas of gene regulation across 21 adult human tissues
K. Fan, A. Sule, A. Guillaumet-Adkins +10
bioRxiv25 Sep 2026PreprintCC BY-NC-NDFull textGenetics & genomicsbioRxiv

Cite

BibTeX RIS

BibTeX

@misc{zheng2026muss,
  title = {{MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation}},
  author = {Kangjie Zheng and Amirhossein Vahidi and Liying Jin and Joseph Clarke and Vijaya Baskar MS and Connor Rogerson and Mengji Zhang and Muzlifah Haniffa and Anshul Kundaje and Mo Lotfollahi},
  howpublished = {arXiv},
  year = {2026},
  month = nov,
  doi = {10.48550/arxiv.2611.08086},
  eprint = {2611.08086},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.08086},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation
AU  - Kangjie Zheng
AU  - Amirhossein Vahidi
AU  - Liying Jin
AU  - Joseph Clarke
AU  - Vijaya Baskar MS
AU  - Connor Rogerson
AU  - Mengji Zhang
AU  - Muzlifah Haniffa
AU  - Anshul Kundaje
AU  - Mo Lotfollahi
PY  - 2026
DA  - 2026/11/05
PB  - arXiv
DO  - 10.48550/arxiv.2611.08086
UR  - https://arxiv.org/abs/2611.08086
AB  - Modeling gene regulation in single cells requires jointly representing DNA sequence, chromatin accessibility, and gene expression across regulatory scales. Sequence-based models capture DNA-level regulatory grammar but often lack cell-level resolution, whereas many single-cell omics models represent genes and peaks as fixed-vocabulary tokens without their underlying sequence. Here, we present MUSS, a sequence-based multi-scale model for single-cell gene regulation that preserves both DNA-level and cell-level resolution. MUSS derives peak- and gene-level representations from DNA using a frozen pre-trained sequence encoder and integrates them through peak-peak self-attention and distance-biased peak-gene interactions. Peak masked language modeling followed by training on paired scATAC-seq and scRNA-seq aligns these representations with cell-specific regulatory states. Across gene expression prediction, enhancer-gene linking, motif-guided regulatory prioritization, and zero-shot cell clustering, MUSS achieves strong performance against sequence-based and single-cell omics baselines. It further predicts expression for genes and peaks unseen during task training. These results support sequence-grounded, multi-scale representations for decoding cell-type-specific regulatory programs. Code, datasets, and model checkpoints are available at https://github.com/Lotfollahi-lab/MUSS_neurips26.
ER  -

Abstract

Related

Cite