NeRV-3D-DC: A Nonlinear Dimensionality Reduction Visualization Method for 3D Chromosome Structure Reconstruction with High Resolution Hi-C Data

Abstract

Chromosome structure plays an essential role in exploring the relationship between chromosome structure and gene transcription. In recent years, High-resolution Hi-C data allows us to explore the chromosome structures in 3D space. However, most of the existing chromosome structure reconstruction methods can not realize the 3D structure reconstruction at a high resolution (for example, 5 Kilobase (KB)). In this paper, we first proposed the NeRV-3D to reconstruct the 3D chromosome structure at a low resolution based on the nonlinear dimensionality reduction visualization algorithm. Then, we proposed the NeRV-3D-DC to reconstruct and visualize the 3D chromosome structure at a high resolution based on the divide-and-conquer method. Results show that NeRV-3D and NeRV-3DDC perform better than existing methods by comparing the 3D visualization effects and evaluation metrics on simulated and real Hi-C datasets. Implementation of NeRV-3D-DC is available on https://githuh.com/ghaiyan/NeRV-3D-DC