Chromatin interactions play significant roles in regulating gene expression. To date, the understanding of the determinants of chromatin interactions remains elusive. Identification of chromatin interaction biomarkers in normal and cancer samples is often a laborious and expensive process, as large amounts of chromatin interaction samples are required. Here, research team led by Dr. Melissa Fullwood, Principal Investigator at the Cancer Science Institute of Singapore (CSI Singapore) and Nanyang Assistant Professor at Nanyang Technological University, has found a method to bridge the research gaps.

In the study recently published in prestigious scientific journal, Genome Biology, in August 2021, Dr. Fullwood and her team members including first authors, Drs. Cao Fan, Zhang Yu and Cai Yi Chao, uncovered that chromatin interactions can be predicted from DNA sequences by leveraging artificial intelligence (AI). The study was conducted in collaboration with Associate Professor Kwoh Chee Keong from the School of Computer Science and Engineering, Nanyang Technological University, Professor Vinay Tergaonkar at the Institute of Molecular and Cell Biology (IMCB), A*STAR, Department of Pathology and Department of Biochemistry, National University of Singapore, and Professor Chng Wee Joo from Cancer Science Institute of Singapore, National University Cancer Institute of Singapore and National University of Singapore.

Previous machine learning approaches that have been developed to detect chromatin interactions suffered from the lack of replicability, with some designed only for specific types of chromatin interactions or were limited to small regions of the genome. Motivated by the lack of effective machine learning approaches, the team developed a computational method, chromatin interaction neural network (ChINN), to predict chromatin interactions between open chromatin regions in the human genome, using DNA sequences and distances as features. This machine learning approach may be able to potentially reduce the number of samples required to identify chromatin interaction biomarkers. As such, the ChINN framework may be useful in the future to study chromatin interactions in large cohorts of clinical samples and identify chromatin interactions-based biomarkers, especially when biological materials are limited. More importantly, the identification of these biomarkers can be used to distinguish between different subtypes of cancer, which will pave the way for the development of precise therapies for different cancer subtypes.

Moving forward, the team hopes to apply the ChINN framework to a large cohort of clinical samples, and further delve into the consequences of DNA mutations or structural variants on chromatin interactions.