Abstract
Chromatin conformation plays an important role in a variety of genomic processes. Hi-C is one of the most popular assays for inspecting chromatin conformation. However, the utility of Hi-C contact maps is bottlenecked by resolution. Here we present VEHiCLE, a deep learning algorithm for resolution enhancement of Hi-C contact data. VEHiCLE utilises a variational autoencoder and adversarial training strategy equipped with four loss functions (adversarial loss, variational loss, chromosome topology-inspired insulation loss, and mean square error loss) to enhance contact maps, making them more viable for downstream analysis. VEHiCLE expands previous efforts at Hi-C super resolution by providing novel insight into the biologically meaningful and human interpretable feature extraction. Using a deep variational autoencoder, VEHiCLE provides a user tunable, full generative model for generating synthetic Hi-C data while also providing state-of-the-art results in enhancement of Hi-C data across multiple metrics.
Highlights
Chromatin conformation plays an important role in a variety of genomic processes
Because all latent vector variables fall within Gaussians centered around 0, most vectors near the center of these Gaussians can be successfully decoded into Hi-C space, resulting in a generative model for Hi-C data
The result is a function mapping a profile of principal component analysis (PCA) values to a 2.57 Mb × 2.57 Mb block of Hi-C data
Summary
Chromatin conformation plays an important role in a variety of genomic processes. Hi-C is one of the most popular assays for inspecting chromatin conformation. We present VEHiCLE, a deep learning algorithm for resolution enhancement of Hi-C contact data. One method to address this issue is to run additional Hi-C experiments, because of experimental costs this is not always a feasible solution To solve this problem previous groups have utilized methods from the field of Image super-resolution to improve Hi-C contact matrix resolution. The first of these networks was HiCPlus[5], a simple neural network optimized using mean squared error (mse). VEHiCLE’s decoder network is engineered to provide an easy to use generative model for Hi-C data generation which smoothly maps user tunable, low dimensional vectors to Hi-C contact maps independent of any low sampled input.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
