AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

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

Kangjie Zheng, Amirhossein Vahidi, Liying Jin, Joseph Clarke, Vijaya Baskar, Connor Rogerson, mj zhang, Muzlifah Haniffa, Anshul Kundaje, Mohammad Lotfollahi

NeurIPS 2026 · Architectures and efficiency › Attention and transformers · approach: State space and recurrent sequence models · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes MUSS, a multi-scale, sequence-based model that jointly represents DNA sequence, chromatin accessibility and gene expression for single-cell gene regulation. (Based on abstract only.)

Problem

Modeling single-cell gene regulation requires jointly representing DNA sequence, chromatin accessibility, and gene expression across regulatory scales; sequence models lack cell-level resolution while single-cell omics models treat genes and peaks as fixed vocabulary tokens.

Approach

Unifies a sequence-based view with cell-state modeling by treating genes and peaks as sequences rather than fixed vocabulary tokens, across multiple regulatory scales. (Abstract only; details unavailable.)

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add frozen DNA sequence embeddings (vs random) (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.196 -> 0.745Improved
add cross-scale peak-gene interaction (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.009 -> 0.745Improved
add RBF distance bias (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.707 -> 0.745Improved
add peak RoPE (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.725 -> 0.745Improved
add distance reweighting (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.736 -> 0.745Improved
add peak MLM pretraining stage (ablation)MUSS · BMMC gene expression ablation, unseen sequencePCC: 0.734 -> 0.745Improved
add peak MLM pretraining stage (ablation)MUSS · zero-shot clustering / enhancer linkingARI / retrievalImproved
add cross-scale interaction (ablation)MUSS · BMMC gene expression ablation, seen sequencePCC: 0.967 -> 0.969No clear effect
use AlphaGenome or Borzoi instead of NT-v3 encoder (ablation)MUSS · unseen-sequence PCC (BMMC)PCC: 0.452 -> 0.745Improved
use AlphaGenome instead of Borzoi encoder (ablation)MUSS · unseen-sequence PCCPCC: BMMC 0.695 -> 0.745; PBMC10K 0.698 -> 0.684Mixed
add cell-contextualized MUSS over underlying sequence encoderAlphaGenome · PBMC169K clusteringARI: 0.1032 -> 0.5377Improved
use MUSS in place of scoobysequence-based predictors · unseen-sequence pseudobulk PCCPCC: 0.533 -> 0.684 (PBMC10K)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy