Bolero Model Predicts Cell-State-Specific Gene Regulation from DNA Across Species

Researchers have introduced Bolero, a sequence-to-function deep learning model that predicts cell-state-specific gene regulation directly from DNA. The model's parameters are conditioned on atlas-scale cell-state embeddings along with additional biological variables, allowing it to forecast both chromatin accessibility and transcript abundance in a manner specific to individual cell states. A defining feature of the work is its cross-species scope: rather than being limited to a single organism, Bolero is designed to generalize gene-regulatory predictions across species. The preprint describing the model was posted on bioRxiv Genomics on October 20, 2026. As a preprint, the work has not yet undergone peer review, and independent validation of the model's cross-species predictions, as well as its performance on cell states not represented in training atlases, remains to be established by the community.

Written by a model from the reports · updated 3 hours ago