Composite Graph Neural Networks for Molecular Property Prediction

Abstract

Graph Neural Networks have proven to be very valuable models for the solution of a wide variety of problems on molecular graphs, as well as in many other research fields involving graph-structured data. Molecules are heterogeneous graphs composed of atoms of different species. Composite graph neural networks process heterogeneous graphs with multiple-state-updating networks, each one dedicated to a particular node type. This approach allows for the extraction of information from s graph more efficiently than standard graph neural networks that distinguish node types through a one-hot encoded type of vector. We carried out extensive experimentation on eight molecular graph datasets and on a large number of both classification and regression tasks. The results we obtained clearly show that composite graph neural networks are far more efficient in this setting than standard graph neural networks.

Keywords: artificial intelligence; composite graph neural networks; deep learning; graph neural networks; molecular graphs; molecular property prediction; open graph benchmark.

Figure 1

The graph neural network model places a copy of the state-updating network over every node of each input graph. After a fixed number of “message-passing” iterations, in which the state of every node is updated based on its previous state, the previous states of its neighbors and the node and edge labels as output are calculated based on the final state. The state-updating network copies share their weights and can be seen as twin building blocks that compose the adaptive architecture of the GNN, replicating the structure of the input graph.