neural-network-MPNN-for-molecular-property-prediction

An Implementation of a type of graph neural network (GNN) known as message passing neural network (MPNN) to predict graph properties. Specifically, we will implement an MPNN to predict a molecular property known as blood-brain barrier permeability (BBBP).

As molecules are naturally represented as an undirected graph G = (V, E), where V is a set or vertices (nodes; atoms) and E a set of edges (bonds), GNNs (such as MPNN) are proving to be a useful method for predicting molecular properties.

Sources and References

• Keras: MPNN example (Kinda depracated)
• Neural Message Passing for Quantum Chemistry
• MPNNModel
• Graph Neural Networks: A Review of Methods and Applications