AI Discovery Reveals DNA Is More Accessible Inside Cells Than Previously Thought

Scientists at Gladstone Institutes and the Arc Institute have used a new AI-powered computational method to challenge a long-standing assumption about how DNA is packaged inside cells. The findings, published in the journal Nature, suggest that genetic material is far more dynamic and accessible than previously believed. For decades, researchers thought DNA wrapped around nucleosome proteins was essentially locked away, with only unwrapped sections available for cellular processes. The new study reveals that most nucleosomes contain partially accessible DNA sections rather than being fully wound up and inactive. The research team, led by Gladstone Investigator Vijay Ramani and Arc Institute Investigator Hani Goodarzi, developed a tool called IDLI that builds on their earlier SAMOSA technology. IDLI uses an AI model trained to recognize subtle structural differences within nucleosomes by scanning sequencing data in two dimensions, both along the DNA fiber and within each nucleosome itself. When the scientists analyzed chromatin from mouse embryonic stem cells, they found that more than 85 percent of nucleosomes showed some degree of distortion, indicating sections of partially exposed DNA. These distortions were not random. The team identified 14 distinct structural states of nucleosomes, each associated with different levels of gene activity. The same patterns appeared in human stem cells being converted into liver-like cells and in liver cells taken directly from mice. The researchers also demonstrated that transcription factors, proteins responsible for turning genes on and off, directly shape nucleosome structures. When two specific transcription factors were removed from cells, the distortion patterns shifted predictably, suggesting these proteins force nucleosomes to either stay open or remain locked. Ramani noted that the 14 nucleosome states could serve as a readout for subtle genetic shifts involved in complex diseases like cancer and neurodegeneration, where small changes across many genes contribute to illness. The team also sees potential applications in aging research, as chromatin structure changes in predictable ways as cells age and some of those changes appear reversible.