AI and the World's Hard Problems · Life sciences · Nature (computational biology) · 2026-10-22
Predicting transcriptional burst kinetics from histone modifications with deep learning
Transcription is inherently bursty, but it has been difficult to quantify at genome scale how histone-modification landscapes encode burst kinetics. A deep learning model called DeepBurst predicts gene-specific burst-frequency and burst-size regimes from histone-modification profiles around transcription start sites, showing consistent performance across three human cell lines and strong cross-cell-type generalization. Model interpretation indicates burst frequency benefits from broader chromatin context while burst size depends more on promoter-proximal signals, and in silico perturbations identify histone-profile changes predicted to shift bursting regimes.