Add the deepFRI neural network to the Rosetta tensorflow_graphs repo

README.md

@@ -6,9 +6,14 @@ A submodule containing Tensorflow models for various machine learning-based prot
 
 * Created as "trRosetta" repo on 21 Jan 2021 by Vikram K. Mulligan, Flatiron Institute ([email protected]).
 * Renamed to "tensorflow\_graphs" repo on 22 Jan 2021.
+* VKM added deepFRI neural network on 27 April 2021.
 
 ## Contents
 
+### deepFRI
+
+A neural network combining an LSTM (long short-term memory) model with a deep graph convolutional neural network to predict protein function from sequence and structure.  Sequence is processed by the LSTM language model, and structure by the deep graph convolutional neural network.
+
 ### trRosetta
 
 A machine learning model that takes an input multiple sequence alignment and produces an output estimate of pairwise residue distances and angles, useful for protein structure prediction.