# arXiv cs.AI: MUSS, a Multi-scale and Sequence-based Model for Single-cell Gene Regulation [Canonical reading](https://escapemeridian.com/readings/arxiv-cs-ai-muss-a-multi-scale-and-sequence-based-model-for-e8c381e7) The authors present MUSS, a model designed to represent DNA sequence, chromatin accessibility and gene expression together while retaining both DNA-level and cell-level resolution. They contrast it with sequence-based models that often lack cell-level resolution and single-cell omics models that represent genes and peaks as fixed-vocabulary tokens without their underlying sequence. MUSS derives peak- and gene-level representations from DNA using a frozen pre-trained sequence encoder, then integrates them through peak-peak self-attention and distance-biased peak-gene interactions. The authors train it first with peak masked language modeling and then on paired scATAC-seq and scRNA-seq to align its representations with cell-specific regulatory states. The authors report strong performance against sequence-based and single-cell omics baselines across gene expression prediction, enhancer-gene linking, motif-guided regulatory prioritization and zero-shot cell clustering. They also report predictions for genes and peaks unseen during task training. This makes the model’s ability to carry sequence-derived representations to unseen genes and peaks a central part of the paper’s claim. The publisher-page abstract does not give quantitative results or name the baselines, so it does not show how large or consistent the reported gains are. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T04:50:38.407Z. First observed: 2026-11-06T05:10:32.414Z. Retrieved: 2026-11-08T04:49:40.570Z. - [MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation](https://arxiv.org/abs/2611.08086) ## Limits The available evidence is the publisher-page abstract; it does not provide quantitative results or identify the baselines used for the reported comparisons. The abstract does not establish how well the reported results transfer beyond the described tasks. ## Open question Which baselines and quantitative results support the authors’ reported strong performance across the four tasks? ## Source-stated The authors present MUSS as a sequence-based multi-scale model intended to retain DNA-level and cell-level resolution. > MUSS, a sequence-based multi-scale model for single-cell gene regulation that preserves both DNA-level and cell-level resolution. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated MUSS derives peak- and gene-level representations from DNA and integrates them through peak-peak self-attention and distance-biased peak-gene interactions. > 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. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated The authors report strong performance against sequence-based and single-cell omics baselines across four named tasks. > 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. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated The authors report predictions for genes and peaks that were unseen during task training. > It further predicts expression for genes and peaks unseen during task training. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source and contributor context Zheng, Kangjie, Vahidi, Amirhossein, Jin, Liying, Clarke, Joseph, MS, Vijaya Baskar, Rogerson, Connor, Zhang, Mengji, Haniffa, Muzlifah, Kundaje, Anshul, Lotfollahi, Mo · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Pre-training A model's initial training on a large dataset, before further adaptation for particular uses. - [Definition reference](https://developers.google.com/machine-learning/glossary#pre-training) ## Zero-shot Doing a task without worked examples (zero-shot) or with only a few examples given in the prompt (few-shot). - [Definition reference](https://developers.google.com/machine-learning/glossary#zero-shot-prompting) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 5d8f3e89-ed85-4759-8bcd-0f97b327ddab. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T04:50:38.407Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: b00c01ce4276826bf78557561f679c172115fddb5f294d1860c0017c43d7c397. Source text hash: aebf808be2db4e0e0d70e20078ce329a2dd566bc9cdafbe996bf6562f72ee74e. ## Revision and source-check history ### Version 1 — 2026-11-08T04:50:38.407Z The authors present MUSS, a model designed to represent DNA sequence, chromatin accessibility and gene expression together while retaining both DNA-level and cell-level resolution. They contrast it with sequence-based models that often lack cell-level resolution and single-cell omics models that represent genes and peaks as fixed-vocabulary tokens without their underlying sequence. MUSS derives peak- and gene-level representations from DNA using a frozen pre-trained sequence encoder, then integrates them through peak-peak self-attention and distance-biased peak-gene interactions. The authors train it first with peak masked language modeling and then on paired scATAC-seq and scRNA-seq to align its representations with cell-specific regulatory states. The authors report strong performance against sequence-based and single-cell omics baselines across gene expression prediction, enhancer-gene linking, motif-guided regulatory prioritization and zero-shot cell clustering. They also report predictions for genes and peaks unseen during task training. This makes the model’s ability to carry sequence-derived representations to unseen genes and peaks a central part of the paper’s claim. The publisher-page abstract does not give quantitative results or name the baselines, so it does not show how large or consistent the reported gains are. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T04:50:38.407Z. First observed: 2026-11-06T05:10:32.414Z. Retrieved: 2026-11-08T04:49:40.570Z. - [MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation](https://arxiv.org/abs/2611.08086) ## Limits The available evidence is the publisher-page abstract; it does not provide quantitative results or identify the baselines used for the reported comparisons. The abstract does not establish how well the reported results transfer beyond the described tasks. ## Open question Which baselines and quantitative results support the authors’ reported strong performance across the four tasks? ## Source-stated The authors present MUSS as a sequence-based multi-scale model intended to retain DNA-level and cell-level resolution. > MUSS, a sequence-based multi-scale model for single-cell gene regulation that preserves both DNA-level and cell-level resolution. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated MUSS derives peak- and gene-level representations from DNA and integrates them through peak-peak self-attention and distance-biased peak-gene interactions. > 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. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated The authors report strong performance against sequence-based and single-cell omics baselines across four named tasks. > 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. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source-stated The authors report predictions for genes and peaks that were unseen during task training. > It further predicts expression for genes and peaks unseen during task training. - [Supporting source](https://arxiv.org/abs/2611.08086) ## Source and contributor context Zheng, Kangjie, Vahidi, Amirhossein, Jin, Liying, Clarke, Joseph, MS, Vijaya Baskar, Rogerson, Connor, Zhang, Mengji, Haniffa, Muzlifah, Kundaje, Anshul, Lotfollahi, Mo · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Pre-training A model's initial training on a large dataset, before further adaptation for particular uses. - [Definition reference](https://developers.google.com/machine-learning/glossary#pre-training) ## Zero-shot Doing a task without worked examples (zero-shot) or with only a few examples given in the prompt (few-shot). - [Definition reference](https://developers.google.com/machine-learning/glossary#zero-shot-prompting) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 5d8f3e89-ed85-4759-8bcd-0f97b327ddab. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T04:50:38.407Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: b00c01ce4276826bf78557561f679c172115fddb5f294d1860c0017c43d7c397. Source text hash: aebf808be2db4e0e0d70e20078ce329a2dd566bc9cdafbe996bf6562f72ee74e. ### uncertain — 2026-11-08T04:50:38.407Z — anthropic/claude-opus-5.5 Publication: 5d8f3e89-ed85-4759-8bcd-0f97b327ddab attribution — pass: The reading consistently attributes the model, method and performance claims to the paper's authors, who are listed on the arXiv page. Results are framed as what the authors report, not as established fact. > Authors:Kangjie Zheng, Amirhossein Vahidi, Liying Jin, Joseph Clarke, Vijaya Baskar MS, Connor Rogerson, Mengji Zhang, Muzlifah Haniffa, Anshul Kundaje, Mo Lotfollahi summary — uncertain: The summary is faithful, but a reader without a genomics background cannot follow it. It never says plainly what gene regulation is or why modeling it in single cells matters. Central terms are used without explanation: chromatin accessibility, peaks, scATAC-seq, scRNA-seq, fixed-vocabulary tokens, frozen pre-trained sequence encoder, masked language modeling and zero-shot cell clustering. Tooltips may cover some of these, but 'peaks' carries much of the argument and needs at least a brief plain-language gloss. Smallest repair: add one opening sentence on what the model is trying to predict, and gloss or tooltip 'peaks' and the assay names. The title and question are fine. > Modeling gene regulation in single cells requires jointly representing DNA sequence, chromatin accessibility, and gene expression across regulatory scales. date — pass: The originalPublished value is 2026-11-05. This matches the v1 submission date on the arXiv page, not the later feed date of 2026-11-06. > \[Submitted on 5 Nov 2026\] support — uncertain: All four listed claims are directly supported, and the limitations accurately note that no numbers or named baselines are given. However, one sentence goes beyond the source: 'This makes the model's ability to carry sequence-derived representations to unseen genes and peaks a central part of the paper's claim.' This is an unlabeled interpretation. The abstract presents unseen-gene prediction as an additional result ('further'). Its stated conclusion is broader: it supports sequence-grounded, multi-scale representations. Smallest repair: label the sentence as Escape Meridian interpretation, or remove 'central'. > It further predicts expression for genes and peaks unseen during task training. media — pass: The selected media is the verified, direct arXiv abstract page for this paper. It is the most direct way for readers to reach the source. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/e8c381e71cde3ef7bb1754aaefa1fb73e4a398ccfb9bc400de74f78513845e3f/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/arxiv-cs-ai-muss-a-multi-scale-and-sequence-based-model-for-e8c381e7.json)