Recapitulation of patient-specific 3D chromatin conformation using machine learning

• DOI: 10.1016/j.crmeth.2023.100578
• Publication Date: 2023-09-05T14:36:53Z
• Authors: Duo Xu; Andre Neil Forbes; Sandra Cohen; Ann Palladino; Tatiana Karadimit...; Ekta Khurana
• Keywords: Chromatin immunoprecipitation; Chromosome conformation capture; Estrogen receptor alpha; ChIP-sequencing

Regulatory networks containing enhancer-gene edges define cellular states. Multiple efforts have revealed these for reference tissues and cell lines by integrating multi-omics data. However, the methods developed cannot be applied large patient cohorts due to infeasibility of chromatin immunoprecipitation sequencing (ChIP-seq) limited biopsy material. We trained machine-learning models using interaction analysis with paired-end tag (ChIA-PET) high-throughput chromosome conformation capture combined (HiChIP) data that can predict connections only assay transposase-accessible (ATAC-seq) RNA-seq as input, which generated from biopsies. Our method overcomes limitations correlation-based approaches distinguish between distinct target genes given enhancers or active vs. poised states in different samples, a hallmark network rewiring cancer. Application our model on 371 samples across 22 cancer types 1,780 602 genes. Using CRISPR interference (CRISPRi), we validated predicted regulate ESR1 estrogen receptor (ER)+ breast A1CF liver hepatocellular carcinoma.