DeepMRF is a hybrid model of deep neural network and Markov Random Field for protein design. In particular, on the one hand, we use Markov Random Filed to capture the very long-range interaction and spatial information among different sites of protein. On the other hand, we use deep learning models to capture the underling complex relationship between protein sequences and structures.
We are busy in other projects and will write up the manscript in near future.