A new model for understanding how single cells regulate genes
MUSS uses DNA sequences to represent gene regulation across different scales.
Researchers created MUSS, a multi-scale and sequence-based model. It models single-cell gene regulation by representing DNA sequence, chromatin accessibility, and gene expression.
The model uses a frozen pre-trained sequence encoder to derive peak- and gene-level representations from DNA. It integrates these through peak-peak self-attention and distance-biased peak-gene interactions.
Peak masked language modeling is used. This is followed by training on paired scATAC-seq and scRNA-seq to align representations with cell-specific regulatory states.
MUSS performs well in gene expression prediction and enhancer-gene linking. It also works for motif-guided regulatory prioritization and zero-shot cell clustering.
The model 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.
Key facts
- MUSS preserves both DNA-level and cell-level resolution.
- The model achieves strong performance against sequence-based and single-cell omics baselines.
- MUSS predicts expression for genes and peaks unseen during task training.