The hierarchical 3D conformation of the genome plays an essential role in gene regulation and proper cell function. With development of the high chromosome conformation capturing technique (Hi-C), researches have revealed the structure of chromatin in various cell types. The Hi-C assay results in a contact map, i.e. a matrix representative of the 3D contacts along the linear chain of DNA, from which various scale-dependent characteristic structures such as chromosomal compartments (100kbp to Mbp) can be deduced. Hi-C has advanced our understanding of genome architecture tremendously. However due to its high cost and time expenditure, increasing effort is being made to derive predictive computational models. Motivated by the field of nonlinear dynamics, in particular chaos theory, we present DNA-DDA, an approach adapted from a time analysis technique, delay differential analysis (DDA), that predicts genome wide contact maps and associated A/B compartments solely from the reference sequence. We hypothesize that sequences close in 3D space will share certain nonlinear dynamical properties to which we can gain access to by mapping the sequences into whats known as an embedding space. Specifically, we used a mere 20Mbp long region on chromosome 22 to generate sparse models for 4 different cell types and achieved exceptional classification performance across all remaining human autosomes. Our approach has the potential to become a powerful tool for studying the biological mechanisms underlying genome folding as well as for modeling the impacts of genetic variation on 3D structure which has been associated with a wide range of diseases.
|Publication status||Published - Sep 2022|
|Event||21st European Conference on Computational Biology - Sitges, Barcelona, Spain|
Duration: 12 Sep 2022 → 21 Sep 2022
|Conference||21st European Conference on Computational Biology|
|Period||12/09/22 → 21/09/22|