A comprehensive map of the sequence basis of transcription initiation in humans

Deep learning models trained on over 200 human transcription initiation profiling experiments accurately predict initiation patterns from local DNA sequence and decode sequence determinants.

• Why it matters: Transcription initiation marks active enhancers and promoters, but linking initiation patterns to DNA sequence across many organs and cancer contexts remains difficult.
• What they did: The study built an atlas of efficient, interpretable deep learning models using more than 200 experiments spanning major human organs and many cancer types, then tested sequence-based prediction and decoding.
• The result: The models’ accurate sequence-based predictions and decoded determinants across diverse cellular contexts enable interpretable mapping of transcription initiation’s sequence basis in humans.

AI summary of the abstract · may contain errors · not medical advice · read the paper.