AI4Loop: an Artificial Intelligence Framework Reveals Increased 3D Chromatin Interactions and Therapeutic Vulnerabilities across 12,000 Cancer Samples
Dao, F.; Lebeau, B.; Amuley, P.; Tan, W. K.; Li, X.; Goh, B. C.; Chng, W. J.; Kwoh, C. K.; Lin, H.; Lyu, H.; Fullwood, M. J.
Show abstract
Three-dimensional chromatin interactions shape gene regulation, but their large-scale analysis remains limited by the cost and complexity of experimental assays. Here we present AI4Loop, a deep learning framework that infers genome-wide gene-centered chromatin interaction networks directly from RNA-seq data. Across multiple cell types, AI4Loop recovered interaction patterns consistent with clinical samples and orthogonal chromatin conformation datasets. Applied to 12,347 transcriptomes from 32 cancer types, AI4Loop revealed pervasive increases in gene-centered chromatin interactions in tumors, particularly at oncogene-associated loci. These inferred interaction networks outperformed gene expression alone in cancer classification. Integration with more than 50,000 drug-treated transcriptomes identified compounds predicted to reverse cancer-associated interaction gains. Hi-C experiments confirmed that the oxazolidinone antibiotics eperezolid and radezolid reduce breast cancer-gain chromatin interactions. Together, these results identify increased gene-centered chromatin interactions as a pan-cancer feature and provide a scalable strategy for linking 3D genome dysregulation to therapeutic vulnerabilities.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Sequence-based modeling of genome 3D architecture from kilobase to chromosome-scale 95%
- Dynamic network-guided CRISPRi screen reveals CTCF loop-constrained nonlinear enhancer-gene regulatory activity in cell state transitions 94%
- Benchmarking of deep neural networks for predicting personal gene expression from DNA sequence highlights shortcomings 94%
Similar papers in this journal
- An interpretable bimodal neural network characterizes the sequence and preexisting chromatin predictors of induced TF binding 95%
- CREaTor: zero-shot cis-regulatory pattern modeling with attention mechanisms 95%
- Geometry aware graph attention networks to explain single-cell chromatin state and gene expression 95%
Similar papers in this journal
- NanoLoop: A deep learning framework leveraging Nanopore sequencing for chromatin loop prediction 95%
- Hierarchical prediction and perturbation of chromatin organization reveal how loop domains mediate higher-order architectures 95%
- Unraveling the molecular mechanisms underlying spontaneous multipolar mitosis through CIN-seq 94%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.