HiCVAE: A Variational Auto-Encoder Framework for Simulating Hi-C Data

Hi-C technology generates substantial sequencing data, which is further complicated by challenges such as the sparsity of samples, spurious associations, and noise. These factors contribute to discrepancies between the sequencing data and the actual structural organization. Moreover, the limited available data samples aggravate the issue. This paper presents a novel method for generating simulated Hi-C data by implementing the variational auto-encoder model. The resultant Hi-C interaction matrix exhibits structural similarity to real data, enabling validation of Hi-C analysis techniques and software, completion of data voids, and streamlining of hypothesis verification experiments.