A new artificial intelligence tool developed at the University of Missouri offers a more precise way to predict the three-dimensional structure of chromosomes in individual cells, which helps overcome a key limitation in genome organization research.
Traditional chromosome conformation capture methods like Hi-C often rely on population-averaged data, making it difficult to distinguish cell-specific folding patterns. The new AI-based model, created by graduate student Yanli Wang and electrical engineering professor Jianlin Cheng, was built to work directly with sparse, noisy single-cell data. This allows it to infer 3D chromosome structure even when parts of the dataset are missing.
“This is important because even cells from the same part of the body can have chromosomes folded in very different ways,” Wang said. “That folding controls which genes are turned on or off.”
Chromosome folding regulates gene activity, and misfolding has been linked to diseases including cancer. Understanding these patterns at the single-cell level has remained a challenge, in part due to the difficulty of modeling structural variability and orientation. The new method incorporates rotational invariance and pattern recognition to outperform existing deep learning approaches, showing more than double the accuracy when applied to human single-cell data.
“Every single cell can have a different chromosome structure,” said Cheng. “Our tool helps scientists study those differences in detail, which can lead to new insights into health and disease.” The researchers have made the tool freely available to the scientific community and plan to extend the model to reconstruct entire genome structures at higher resolution.
Publication Details
Yanli Wang, Jianlin Cheng, Reconstructing 3D chromosome structures from single-cell Hi-C data with SO(3)-equivariant graph neural networks, NAR Genomics and Bioinformatics, Volume 7, Issue 1, March 2025, lqaf027, https://doi.org/10.1093/nargab/lqaf027
Traditional chromosome conformation capture methods like Hi-C often rely on population-averaged data, making it difficult to distinguish cell-specific folding patterns. The new AI-based model, created by graduate student Yanli Wang and electrical engineering professor Jianlin Cheng, was built to work directly with sparse, noisy single-cell data. This allows it to infer 3D chromosome structure even when parts of the dataset are missing.
“This is important because even cells from the same part of the body can have chromosomes folded in very different ways,” Wang said. “That folding controls which genes are turned on or off.”
Chromosome folding regulates gene activity, and misfolding has been linked to diseases including cancer. Understanding these patterns at the single-cell level has remained a challenge, in part due to the difficulty of modeling structural variability and orientation. The new method incorporates rotational invariance and pattern recognition to outperform existing deep learning approaches, showing more than double the accuracy when applied to human single-cell data.
“Every single cell can have a different chromosome structure,” said Cheng. “Our tool helps scientists study those differences in detail, which can lead to new insights into health and disease.” The researchers have made the tool freely available to the scientific community and plan to extend the model to reconstruct entire genome structures at higher resolution.
Publication Details
Yanli Wang, Jianlin Cheng, Reconstructing 3D chromosome structures from single-cell Hi-C data with SO(3)-equivariant graph neural networks, NAR Genomics and Bioinformatics, Volume 7, Issue 1, March 2025, lqaf027, https://doi.org/10.1093/nargab/lqaf027